SlideShare une entreprise Scribd logo
1  sur  29
BlockMed: Open Protocol for Medical IoT
Wayne Chung, Andrew Lee
Dec. 8th, 2018
Abstrac​t
We present an open protocol for medical IoT devices that connects payer, provider and patients
on a public Ethereum blockchain that allows for trustless, secure smart contract execution. Our
protocol allows for BlockMed token, BMD to serve as a utility token which allows 3 core
innovations: 1) Distributed encryption/decryption of HIPAA protected medical IoT patient data
for prevention of hacking and ransomware type attacks in remote monitoring settings. 2)
Augmented deep learning and blockchain platform to allow faster, more efficient development of
crowdsourced AI algorithms. 3) Machine to machine payment of smart contract parameters from
medical IoT data based on individual and community standards.
1
Contents
I. Introduction…………………………………………………………… 3
II. Current State ………………………………………………………….. 4
III. Enhanced Cybersecurity ………………………………………………. 5
IV. Crowdsourced AI Algorithms ………………………………………….14
V. Machine to Machine Payments ……………………………………….. 20
VI. Token Economics & Governance……………………………………... 23
VII. Timeline ………………………………………………………………. 24
VIII. Summary ……………………………………………………………… 26
IX. Acknowledgements …………………………………………………… 26
X. References …………………………………………………………….. 27
2
I. ​Introduction
The rapid proliferation of medical IoT devices recording patient data remotely outside the
hospital and clinic by connecting to the mobile-cloud is creating increasing cybersecurity
vulnerabilities for the health IT infrastructure. Provider based EHRs are understandably
concerned and reluctant to open their APIs to the myriad of remote monitoring devices
connecting to the mobile-cloud for fear of hacking and ransomware type attacks. Enhanced
cybersecurity of existing health IT and EHRs is top of mind for CIO’s of all major health
organizations. BlockMed proposes an open protocol which allows remote medical IoT data to be
HIPAA compliant on and off the blockchain so there is record of sensor data verified with the
patient, provider and payer. Username-password compromises, as seen in centralized databases
allowing current ransomware attacks, will be more difficult on the distributed platform proposed
here.
The exponential explosion of medical IoT data is becoming a double-edge sword regarding
utility and value created from such data. Providers are increasingly overwhelmed from alarm
fatigue and false positives which generate unnecessary calls, visits, and further diagnostics,
potentially creating more legal liability. The fundamental value of this data resides in the
actionable algorithms created to predict and inform therapeutic action. Creating such actionable
algorithms depends heavily on the quantity and quality of the data, as compared to the entire
body of data from the individual and community.
Remote medical IoT data analysis are currently limited to the number of patients which can
generate that data and the team members within an individual company to create actionable
algorithms from that data. BlockMed proposes a novel open platform whereby such remotely
generated medical IoT data can be crowdsourced out to the BMD community to then develop
superior algorithms, that overcomes standard overfitting bias based on static historical data as
opposed to new continuously generated live data. This deep learning blockchain platform will
3
allow faster, more efficient development of crowdsourced AI algorithms, that can then result in
machine to machine payment of smart contract value from medical IoT data based on real-time
live data as opposed to traditional population health models currently in existence.
II. ​Current State
Blockchain technology applied to medical IoT is still in its infancy. Decentralized distributed
ledger technology is ideal for immutable records of crowdsourced algorithm performance and
machine to machine payments where secure identification and verification of algorithm results is
crucial for widespread adoption. Many other proposed projects identify the EHR and existing
legacy IT systems as starting point for a blockchain based replacement. This rapidly growing
eco-system can accommodate many approaches, but in our opinion the medical IoT space is
relatively new like blockchain--therefore a prime opportunity to rapidly establish a widely
adopted and scalable use case as there is no current EHR or IT legacy system to overcome and
replace.
Other proposed projects attempt to combine predictive algorithms with blockchain for better
disease prediction on existing data sets from current hospital or clinic-based systems. Such
projects have issues of ownership of data, as the existing legacy hospital and clinic systems are
reluctant to enter into profitable agreements to share this data for the proposed types of
blockchain projects. Without this legacy data in actionable formats from current hospital or
clinic-based IT systems, blockchain development of predictive algorithms in these types of
projects may be much slower in getting massive data to scale and yielding superior actionable
results.
Supply chain management of hospital and medical device supplies are other proposed blockchain
projects. Given the complicated logistics involved with large legacy systems, this area is better
suited for large existing IT vendors such as IBM, Microsoft, AWS to offer blockchain services
4
for large companies to deploy--as IBM has already done with the cargo shipping industry.
Simply replicating entire aspects of legacy health IT infrastructures with blockchain is unlikely
to gain rapid adoption and scale if the complexity of legacy systems and lack of data access
continues--as will likely be the case while the legacy incumbents slowly adopt blockchain on
their own.
Areas which fundamentally lends itself to open protocol development have the best chance for
rapid network adoption and true scaling potential for existing healthcare players. BlockMed is
proposing an open protocol which allows medical IoT data largely untethered to current legacy
IT systems, to be rapidly adopted by payers, providers, and patients in a BMD token eco-system
with core innovations in enhanced cybersecurity, crowdsourced AI algorithms and machine to
machine payment.
III. ​Enhanced Cybersecurity
BlockMed will utilize a patented wearable multi-functional sensor​[1]​
to initially provide data for
its proposed open protocol for enhanced cybersecurity of medical IoT. This proprietary sensor
has been developed for over 5 years by Wireless Medical for remote monitoring of cardiac
disease. It produces 9 real-time data streams from a credit card size sensor worn on chest that
sends data via Bluetooth to an iOS app on iPhone, which streams the data to AWS cloud server,
where it is stored and sent back to iOS app for display of data to physician.
For HIPAA compliance, the data streamed from sensor via Bluetooth will be encrypted with
AES-128 protocols. The data streamed from iOS app to web app will be protected by SSL
encryption, while the data stored in AWS cloud server will have dedicated hardware provided for
encrypted data at rest. For sensor’s 9 real-time data streams, no such encryption is necessary as
long as there is no pairing with the HIPAA compliant patient identifiers. The raw sensor data
will be stored on public distributed file storage protocols such as IPFS​[2]​
via BlockMed open
5
source API​[3]​
(See Figure 1). This not only enables other technology and framework such as IoT
Fog/Edge computing framework proposed by Cisco​[4]​
to take advantage of IPFS that has better
performance than others such as Amazon S3​[5]​
, but also shows better performance gain with
IPFS​[6]​
.
Figure 1
The proposed open protocol will utilize a public Ethereum​[7]​
blockchain for the smart contract
execution, while the HIPAA compliant patient identifier data will be stored off-chain such as an
AWS encrypted databases as necessary for proper onboarding of medical sensor usage. The
private keys of public Ethereum blockchain can be kept on mobile app of patient. In case of loss
of mobile app/smart phone, this unique key can have appropriate recovery or replacement
mechanisms provided by BlockMed​[8]​
. Furthermore, the API kit can be extended to support data
encryption on top of IPFS as an extra layer of security to keep data encrypted and give users full
control of their data. In addition to user data, the open protocol also provides a channel to
compute Private Contracts, in other word, confidential code that can be service through off-chain
service provider that provides TEE (Intel SGX) backed compute-nodes​[9][10]​
. In Oasis Lab’s
Ekiden​[11]​
protocol, clients send inputs to confidentiality-preserving smart contracts, which are
executed within a TEE at any compute node. The blockchain stores encrypted contract state.
6
This provides opportunity for regulated industries to utilize the open protocol and compute
public data from IPFS along with private data off-chain with proprietary code. The owner of the
data who holds the Private Key can authorize different service providers with the proper key to
decrypt the data and still track of each service provider on how the authorization are distributed
and applied. Other services such as Enigma​[12]​
can also take advantages to corporate data privacy
off-chain​[13]​
by exchanging the authorized secret key access from user via Ethereum smart
contract while the data are still kept on the public IPFS for further usage. Enigma also opens the
door to multi-party computation, MPC​[14][15]​
as well. This also leaves a trace of who accessed the
data on the public chain as an evidence and audit. (See Figure 2)
Figure 2
The distributed ledger technology of public Ethereum blockchain provides enhanced security
with immutable records of transactions on nodes. As such, it is very difficult to hack in
traditional ways seen in centralized databases such as AWS, often through username-password
compromises seen frequently in ransomware attacks on hospital systems. BlockMed’s proposed
open protocol cooperating with private off-chain, AWS encrypted database for patient, physician
ID, and public IPFS file storage for raw sensor data serves as redundancy for both EHRs’ and
Payers’ centralized databases. (See Figure 3)
7
Figure 3
In cases of ransomware attacks on institutions, utilizing BlockMed’s proposed open protocol, the
distributed and decentralized private keys stored on patients’ mobile apps provides an additional
level of security for the medical IoT data. As a result, there may be little incentive to pay
ransomware demands as the distributed IPFS file storage data can be easily recovered and
duplicated onto legacy systems. (See Figure 4)
Attack Mitigation
The following illustrate how a potential attack scenario by Ransomware or compromising the
server managed by central authority for business usage.
8
Figure 4 - Attack Scenario with Ransomware
According to Figure 4, immutable records and data are already anonymized and de-identified on
IPFS with the list of pointers hosted by one or more Open Source Web Server​[16]​
(via Caching
and Indexing). This means by gaining access to the server or the Open Source Web Server will
limit the chance to correlate and expose patient’s identities based on the randomized UUID.
This Open Source Web Server can be replicated and distributed by anyone anywhere for high
availability, reliability. and performance​[17]​
. The authority can decide how many of the caching
servers they would like to host internally to reach a quorum for reliability in their own
environment, or simply become part of the public Internet that interacts with other alike servers
(however, this is likely to happen due to each authority will like to keep track of a minimal
subset of the global records on IPFS, especially from their own patients). As a result, the victim
authority will be able to perform the following to recover from such attack. (See Figure 5)
1. Setup a new secured server with latest security patch, etc.
2. Migrate the IP address from the old server to the new one
9
a. Existing sensors will retry to register on this ‘new’ server, and afterward, continue
to send data to it. This provides the server a list of existing devices (aka previous
registered patients).
b. In addition to the new incoming data after the service availability is restored,
without the existing sensors sending data, the ‘new’ server can coordinate with
the Open Source Web Server to pre-load all the historical index to fetch the data
from IPFS. This will allow the ‘new’ server to bootstrap the lost historical data,
meanwhile, still receive new (incremental data) from the existing sensors.
3. After a certain period of time, the ‘new’ server will gain its original view prior to the
attack, and hence, can mitigate the attack (e.g. ransomware).
Figure 5 - Recovery Scenario to Restore Data
10
BlockMed’s proposed open protocol allows for developers worldwide to contribute to the
sensors’ raw data storage through the BlockMed token, BMD. Wireless Medical’s proprietary
sensor’s 9 real-time data streams accumulating 24/7 daily quickly adds up to large amounts of
data. Blockchain technology as applied to healthcare IT security is still in its infancy. Two main
issues BlockMed will address along with entire public Ethereum blockchain community include:
1) tradeoffs between maintaining security while increasing transaction speeds to handle
increasingly large amounts of raw data on the public Ethereum blockchain and IPFS file storage
network, (See Figure 6)
Figure 6 - Tradeoff Between Ethereum and IPFS Storage
2) having useful off-chain API’s to interact with public Ethereum blockchain and IPFS file
storage networks in order to handle all the various medical IoT devices, EHR’s, Insurance
Payers, App/Browser eco-systems etc. that will be necessary for future scaling during
mainstream adoption phase.
BlockMed’s internal team and open community developers will use BMD token incentives to
contribute data from Wireless Medical’s multi-functional sensor as well as their own off the shelf
11
sensor data to the distributed IPFS file storage platform. Consumer devices with medical
applications such as Apple Health Kit API, Apple Watch, Fitbit and other app systems such as
Android, and other web browser-based systems can also be tested and developed on BMD open
platform by developers. This growth will be powered by worldwide community of developers
incentivized through BMD token to improve functionality and security of open platform for
medical IoT devices. (Figure 7)
Figure 7 - Multiple Wearable Devices Integrated with Ethereum and IPFS
This platform will initially interface with the existing Swift iOS app, Ruby-Python server and
MongoDB database residing on AWS. (See Figure 8)
12
Figure 8 - Initial Implementation with IPFS and BMD Smart Contract
As improvements are made to the distributed IPFS file storage platform handling the open
communities’ sensor data, consensus mechanisms will be utilized through BMD smart contracts
to incentivize changes that achieve stated goals of increasing speed and efficiency in handling
raw sensor data on distributed IPFS as well as usefulness in interacting with appropriate API’s
for off chain EHR’s such as Epic which is used in Stanford, UCSF hospitals, where initial pilots
will be conducted, and API’s for off chain Insurance Payers such as United Health, Humana, or
others which may enter into agreements during pilot studies.
In Stanford, UCSF pilot studies, the patient and physician ultimately should not notice any
difference in testing the Wireless Medical multi-functional sensor and mobile app with regards to
the data streams collected, between non-blockchain and BMD blockchain versions for enhanced
cybersecurity. (See Figure 9)
13
Figure 9 - Interaction with Ethereum and IPFS are Transparent
After successful pilot studies, emphasis will be on API integration and bringing on commercial
partners onto BlockMed open platform with other hospital systems, EHRs, Payers, and other
medical IoT devices.
IV. ​Crowdsourced AI Algorithms
BlockMed’s 2​nd​
core innovation is that Wireless Medical’s multi-functional proprietary sensor
algorithms can be crowdsourced real-time to open community through use of BMD tokens to
produce Dutch​[18]​
style multi-unit auctions that result in superior machine learning algorithms on
each of the 9 sensor data streams, as well as a combined composite score. This technique can
overcome issues​[19]​
of overfitting algorithms​[20]​
to static historical test data that are poor
performing on newer real-time data streams. This is critical in the medical IoT space, not only
for Wireless Medical but other medical IoT devices, which all constantly produce new 24/7
real-time sensor data under wide variety of variable environmental conditions.
The 9 data streams produced by Wireless Medical’s sensor include: 1) acoustic heart sounds 2)
skin impedance 3) EKG 4) heart rate 5) heart rate variability 6) respiratory rate 7) temperature
8) sleep angle 9) steps walked. Each of these data streams represent physiological states in
14
patient wearing sensor, but exist in noisy remote, non-clinical environment where movement and
other factors can introduce artifacts into data streams. (See Figure 10)
Figure 10 - De-identified Input Stream from IPFS to AI Frameworks
In proposed open protocol, the sensor’s raw data streams, de-identified on distributed IPFS file
storage network, can be made available to community of BMD token holders and developers,
who can utilize various machine learning and deep learning algorithms frameworks such as
TensorFlow, PyTorch, CNTK, and MXNet ​to more efficiently refine accuracy not only on static,
old historical data, but on constantly generated new live data. (See Figure 11)
Figure 11 - General AI Framework Tapping into BMD Protocol
15
A multi-unit Dutch auction can be utilized so algorithms performing better on new data get
incentivized through BMD tokens issued to these developers in regularly scheduled contests​[21]​
.
Developers can also use their own off the shelf consumer devices for commodity data streams
such as HR, HRV, RR, Temp, Sleep Angle and Steps to submit algorithms which can be tested
against other consumer devices as well as Wireless Medical’s proprietary multi-functional
sensor. Complex, unique data streams such as acoustic heart sounds, skin impedance and EKG’s
can only be gathered from Wireless Medical’s proprietary sensor. Initially this data will be
gathered and tested on healthy developers from BlockMed’s internal team with access to
proprietary sensor, but later on sick patients in the UCSF, Stanford clinical studies. These
complex, unique data streams can also be crowdsourced and incentivized through a multi-unit
Dutch auction to BMD developer community just like the commodity data streams.
The Dutch auction’s goal will be to produce algorithm models which accurately predict behavior
of both individual data streams and combined sensor score as related to true physiological state
of patient. For example, raw HR data gathered from Wireless Medical’s multi-functional sensor
can be used to create predictive algorithm models, then tested on Apple Watch’s raw HR data
and vice versa. These models will need to account for occasional falsely high HR’s due to
wearable environment as compared to clinically labelled HR as gold standards. Best models
submitted by BMD token community of developers can be forward tested on wide variety of
other devices new raw HR data for a continuously evolving winning model. Mechanics of
reward amounts, schedule and winners can be governed by existing best practices and modified
by BMD community through consensus. This Dutch auction can be applied to all 9 individual
data streams and combined composite score produced by Wireless Medical’s multi-functional
sensor. (See Figure 12)
16
Figure 12 - Derived and Published Algorithm from Crowdsourced AI gets Rewarded
The BMD crowdsourced machine learning algorithms can also be applied to the mobile data
generated through future voice activated app platform along with camera, video data upon app
interaction by patient. True artificial intelligence through applying deep learning natural
language processing and computer vision techniques to mobile audio​[22]​
and video​[23][24]​
data
streams along with machine generated sensor data and be compared to human labelled clinical
data in clinical studies over a longitudinal time frame of 7-day hospital stay and 30-day
discharge period at home. A combined composite mobile and sensor score of all 9 data streams
equally weighted can be compared to equivalent composite clinical score of 9 physician reported
parameters and compared for correlation and predictive power in the hospital and later in
patient’s home environment. Again, by crowdsourcing to BMD token developer community, the
efficient optimization of superior algorithms—a world’s first true deep learning platform on the
blockchain can be created not only for Wireless Medical’s multifunctional sensor, but ultimately
17
for other medical IoT devices generating massive amounts of 24/7 physiological data real
time​[25]​
.
Figure 13 - Data Published to IPFS and BMD Smart Contract
The BMD token developer community can be incentivized to analyze open source community
data such as Beth Israel—MIT EKG database for arrhythmias, or TI’s database for HR detection
algorithms and efficiently optimize algorithms on both individual and community level.
Ultimately the goal will be to generate supervised, semi-supervised and unsupervised deep
learning algorithms​[26][27]​
that trains on labelled and unlabeled open source community data and
individual medical IoT data to perform accurately on newly generated patient data across wide
variety of settings, as that will prove a most powerful and valuable use case. Unlike previous
and existing legacy systems, this open protocol allows the BMD token developer community to
capture much of this value creation in the deep learning algorithms of newly generated mobile
and sensor data as shown to be useful for the patient and physician.
The de-identified raw sensor data on public distributed IPFS file storage network will always be
kept separate from patient, physician identifiers kept on off-chain, encrypted AWS. ​One
18
possibility is to run Apache Spark​[28]​
on Amazon EMR, and use Terraform​[29]​
(by HashiCorp) to
build a Spark and Apache Zeppelin​[30]​
cluster on Amazon EMR which is HIPAA compliant​[31]​
.
This solution encrypts all data at rest and in-flight, logs all user activities, as well as satisfies
many other aspects of a HIPAA compliant environment. The use of Terraform provides a high
degree of management of Cluster Configuration, Data Accessibility, Scalability, Security, and
Availability. (See Figure 14)
Figure 14 - Running Big Data Applications in HIPAA Environment
As such HIPAA compliance can be met, while allowing open BMD developer community to
rapidly iterate, and test mobile and sensor data with real world clinical inputs to create a deep
learning blockchain platform which can be applied not only to affect various cardiac conditions
such as heart failure, cardiac arrhythmias, post heart surgery management but also other
non-cardiac entities such as pulmonary diseases etc.
19
V. ​Machine to Machine Payments
BlockMed’s 3​rd​
core innovation is to allow machine to machine payments on the open Ethereum
blockchain platform, such that any set of individual and community parameters can be set in the
smart contract as incentives for individual data streams and combined sensor score results.
Current medical IoT devices have to go through large, expensive studies to prove positive
changes in clinical outcomes worthy of reimbursement by Payers. These population-based
studies provide a major hurdle to widespread medical IoT adoption as the desired outcomes
measure results in lengthy binary decisions to reimburse or not based on entire population tested
when certain patients may benefit at individual level. (See ​Figure 15)
Figure 15 – Current Medical IoT Ecosystem
The BMD open platform will allow Payer to incentivize each sensor data stream and combined
score to standards on individual real-time level. For example, HR can be tracked to certain
normal ranges, and deviations from that community standard can be verified on public Ethereum
blockchain platform. Patients’ that have minimal deviations can then have de-identified BMD
token transfers on public Ethereum blockchain that connects to Payer’s off-chain encrypted
account of that individual with resulting co-pays, deductibles etc. available to adjust accordingly.
Such value transfers with BMD tokens real-time from Payer to device at individual level allows
for faster commercial adoption of medical IoT devices as Payers do not necessarily need to wait
until binary outcomes data from large commercial adoption studies when deciding on
reimbursement in remote monitoring setting. Instead, medical IoT devices can be incentivized to
20
join open platform with Payers setting individual standards for data streams regarding BMD
token value transfers. (See ​Figure 16)
Figure 16 - BlockMed Token Ecosystem
Positive reinforcement for clinical compliance as evidenced by medical IoT data that adheres to
community standards can now be transferred through BMD tokens real-time, machine to
machine. Payers that allow select patient community data to be de-identified and made available
to open platform can benefit from crowdsourced modelling of most efficient payment incentives
at individual level, similar to Dutch auction model used for deep learning algorithm development
regarding accuracy of individual and composite data streams. Again, the BMD platform allows
developers who create the most accurate algorithms for Payers to transfer token value can
capture most of the value creation, as there is much potential cost reduction in entire healthcare
system regarding prevention and management of chronic disease in remote monitoring setting.
(See ​Figure 17)
21
Figure 17 - Network Virality Effect and BMD Circulation
The medical IoT space is ideal for machine to machine payments as typical hospital
re-admissions can cost Payer $10-15K, and happen 20-25% of time on all major admission
categories such as heart failure, pneumonia, heart surgeries etc.​[32]​
. In addition, patient
deductibles are increasingly high, such as $3-10K on all types of Payer plans. Crowdsourced
deep learning models which accurately predict individual data streams at individual level, then
tie combined composite score directly to Payer accounts of deductibles and premiums, all on
off-chain, encrypted AWS and public Ethereum blockchain platforms offer unique opportunity to
transform fundamental payment models currently in place primarily for inpatient and clinic
visits. (See ​Figure 18)
22
Figure 18 - Enterprise and Secure Computation in BMD Network
Reimbursement for medical IoT remote monitoring is still very much in its infancy due to lack of
quantifiable data regarding effectiveness of various low-tech disease management approaches as
well as digital therapeutic, medical IoT offerings. BMD’s open platform allows for quantified,
results orientated, machine to machine micro-payments at individual level, based on medical IoT
real-time data streams.
VI. ​Token Economics & Governance
BlockMed will issue its BMD tokens through pre-sale and later it in public sale on rolling basis
until funds deemed sufficient for network development. A trusted public wallet managing the
escrow account with the initial 1B tokens such as CoinBase with secure offline cold storage
23
capabilities will be used to safely handle the exchange of Ethereum, Bitcoin, USD, Euro into
BMD tokens. Appropriate identification provided by third party vendors will be requested so
Know Your Customer (KYC), Anti-Money Laundering (AML) and all other relevant regulations
are complied with. A multi-sig. contract whose keys are held by trusted individuals within
BlockMed will be used to handle all funds. 1 billion total BMD tokens will be issued at initial
exchange rate of 1USD per BMD token. Allocations include: 20% to founding team and
advisors, 30% to BMD Foundation, 50% to public. (See ​Figure 19)
Figure 19 – BlockMed, BMD ERC20 Smart Contract Execution
BlockMed’s initial governance will be guided by founding team, board of directors to set
direction of strategy, including key internal hires and outside engineering design/ development
houses to execute out private and public blockchain platforms in preparation for UCSF, Stanford
clinical pilot and later commercial network Payer partners. To promote a common standard
among the open protocol proposed by BlockMed, decentralized consensus may be utilized with
open governance and voting among the BMD token holders, guided by BlockMed’s founding
24
team. Later a true decentralized autonomous organization (DAO) governance structure may be
considered once the open protocol has reached sufficient community adoption.
Figure 20 - Token Allocation
VII.​ ​Timeline
Jan.--Aug. 2018—Whitepaper Research
May--Dec. 2108—Smart Contract Prototype
Jan.--Dec. 2019—Rolling Raise BMD Tokens
Jan.--Dec. 2019—BlockMed.AI Platform Development
July--Dec. 2019—UCSF, Stanford Pilot Studies
Jan.--Dec. 2020—Payer / Pharma Licensing
25
VIII. ​Summary
Remote medical IoT data has the potential to fundamentally transform the healthcare landscape
with regards to chronic care disease management, keeping patients out of the hospital and
allowing more cost savings to entire system through early prediction and therapeutics.
Unfortunately, the current system depends on large scale trials and pilots with payers and
providers to prove cost savings on community populations before large scale deployment and
reimbursement. BlockMed proposes a novel, secure, deep learning blockchain platform, based
on the BMD token, whereby large population-based community trials are not a necessary
precursor to individual transfer of value real time based on medical IoT data.
Based on public Ethereum smart contract parameters, remote medical IoT data that meets certain
algorithmic criteria can result in real-time machine to machine payments through BMD token.
Such a protocol has potential to transform transmission of value throughout the entire payer,
provider, patient ecosystem for medical IoT, and truly incentivize improved health outcomes and
lower costs among all affected parties.
IX. Acknowledgements
We would like to thank our mentors, advisors, and friends who have provided invaluable advice
on BlockMed. In particular, for educating, reviewing and providing feedback on this work in
specific and blockchain-cryptocurrency world in general.
26
X. References
[1] WMMIP101US; U.S. Application Serial No. 14/050,356; Issue Notification 08-21-18
[2] ​https://github.com/ipfs/papers/raw/master/ipfs-cap2pfs/ipfs-p2p-file-system.pdf​ and J. Benet,
“IPFS - Content Addressed, Versioned, P2P File System,” Protocol Labs, Inc., Tech. Rep., 2014.
[3] ​https://github.com/blcksync/bc-ipfs
[4] F. Bonomi, R. Milito, J. Zhu, and S. Addepalli, “Fog Computing and Its Role in the Internet
of Things,” in ​Proceedings of the First Edition of the MCC Workshop on Mobile Cloud
Computing​, ser. MCC ’12, 2012, pp. 13–16. ​https://github.com/ethereum/wiki/wiki/White-Paper
[5] B. Confais, A. Lebre, and B. Parrein, “Performance Analysis of Object Store Systems in a
Fog/Edge Computing Infrastructures,” in ​IEEE CloudCom​, Luxembourg, Luxembourg, Dec.
2016.
[6] ​https://hal.archives-ouvertes.fr/hal-01483702/document
[7] ​https://github.com/ethereum/wiki/wiki/White-Paper
[8] ​https://github.com/bitcoin/bips/blob/master/bip-0044.mediawiki
[9] Fisch, B., Vinayagamurthy, D., Boneh, D., and Gorbunov, S. Iron: functional encryption
using Intel SGX. In ​Proceedings of the 2017 ACM SIGSAC Conference on Computer and
Communications Security​ (2017), ACM, pp. 765–782
[10] Nayak, K., Fletcher, C., Ren, L., Chandran, N., Lokam, S., Shi, E., and Goyal, V. Hop:
Hardware makes obfuscation practical. In ​24th Annual Network and Distributed System Security
Symposium,​ NDSS (2017)
[11] Ekiden: A Platform for Confidentiality-Preserving, Trustworthy, and Performant Smart
Contract Execution ​https://arxiv.org/abs/1804.05141
[12] ​https://enigma.co/protocol/SoftwareArchitecture.html
[13] Bogdanov, Dan, Sven Laur, and Jan Willemson. ”Sharemind: A framework for fast privacy-
preserving computations.” ​Computer Security-ESORICS​ 2008. Springer Berlin Heidelberg, 2008.
192-206
[14] Yao, Andrew C. ”Protocols for secure computations.” 2013 IEEE ​54th Annual Symposium
on Foundations of Computer Science​. IEEE, 1982
27
[15] Lindell, Y., and Pinkas, B. Secure multiparty computation for privacy- preserving data
mining. ​Journal of Privacy and Confidentiality​ 1, 1 (2009), 5
[16] ​https://en.wikipedia.org/wiki/Comparison_of_web_server_software
[17] Consistency of Data replication protocols in database systems: a review - ​International
Journal on Information Theory (IJIT),​Vol.3, No.4, October 2014
[18] ​Vickrey, W. (1961). Counter speculation, auctions, and competitive sealed tenders. The
Journal of Finance​, 16(1), 8–37. A pathbreaking paper that introduced second price auctions and
performed new analysis of first price
[19] Hardt. Competing in a data science contest without reading the data.
Http://blog.mrtz.org/2015/03/09/competition.html​​
[20]​​
Dwork, Feldman, Hardt, Pitassi, Reingold, Roth. Generalization in Adaptive Data Analysis
and Holdout Reuse.
http://papers.nips.cc/paper/5993-generalization-in-adaptive-data-analysis-and-holdout-reuse.pdf​.
[21] Numerai - ​https://numer.ai/static/media/whitepaper.29bf5a91.pdf
[22]​​
Chan, W.; Jaitly, N.; Le, Q.; and Vinyals, O. 2016. Listen, attend and spell: A neural
network for large vocabulary conversational speech recognition. In ​​Acoustics, Speech and Signal
Processing (ICASSP), 2016 IEEE Inter- national Conference on​​, 4960–4964. IEEE
[23] Esteva, A.; Kuprel, B.; Novoa, R. A.; Ko, J.; Swetter, S. M.; Blau, H. M.; and Thrun, S.
2017. Dermatologist-level classification of skin cancer with deep neural networks. ​​Nature
542(7639):115–118
[24] Gulshan, V.; Peng, L.; Coram, M.; Stumpe, M. C.; Wu, D.; Narayanaswamy, A.;
Venugopalan, S.; Widner, K.; Madams, T.; Cuadros, J.; et al. 2016. Development and validation
of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs.
Jama ​316(22):2402–2410
[25] Tison, G. H.; Singh, A. C.; Ohashi, D. A.; Hsieh, J. T.; Ballinger, B. M.; Olgin, J. E.;
Marcus, G. M.; and Pletcher, M. J. 2017b. Abstract 21042: Cardio- vascular risk stratification
using off-the-shelf wearables and a multi-task deep learning algorithm. ​​Circulation ​​136(Suppl
1):A21042–A21042
28
[26] Dai, A. M., and Le, Q. V. 2015. Semi- supervised sequence learning. In ​​Advances in Neural
Information Processing Systems​​, 3079–3087
[27] LeCun, Y.; Bengio, Y.; and Hinton, G. 2015. Deep learning. ​​Nature ​521(7553):436–444
[28] ​https://spark.apache.org/
[29] ​https://www.terraform.io/
[30] ​https://zeppelin.apache.org/
[31] ​https://aws.amazon.com/compliance/hipaa-eligible-services-reference/
[32] 2010 ​International Cochrane Collaboration​ review of 25 RCT trials: 9,500 CHF patients,
remote monitoring services resulted in 33% drop of mortality. In contrast, nurse monitoring of
heart failure patients by telephone was not shown to reduce mortality, although it was linked to
lower readmission rates. Both approaches cut overall health costs.
29

Contenu connexe

Dernier

Call Girls In Noida 959961⊹3876 Independent Escort Service Noida
Call Girls In Noida 959961⊹3876 Independent Escort Service NoidaCall Girls In Noida 959961⊹3876 Independent Escort Service Noida
Call Girls In Noida 959961⊹3876 Independent Escort Service Noida
dlhescort
 
Call Girls From Pari Chowk Greater Noida ❤️8448577510 ⊹Best Escorts Service I...
Call Girls From Pari Chowk Greater Noida ❤️8448577510 ⊹Best Escorts Service I...Call Girls From Pari Chowk Greater Noida ❤️8448577510 ⊹Best Escorts Service I...
Call Girls From Pari Chowk Greater Noida ❤️8448577510 ⊹Best Escorts Service I...
lizamodels9
 
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
dollysharma2066
 
FULL ENJOY Call Girls In Majnu Ka Tilla, Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Majnu Ka Tilla, Delhi Contact Us 8377877756FULL ENJOY Call Girls In Majnu Ka Tilla, Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Majnu Ka Tilla, Delhi Contact Us 8377877756
dollysharma2066
 
Chandigarh Escorts Service 📞8868886958📞 Just📲 Call Nihal Chandigarh Call Girl...
Chandigarh Escorts Service 📞8868886958📞 Just📲 Call Nihal Chandigarh Call Girl...Chandigarh Escorts Service 📞8868886958📞 Just📲 Call Nihal Chandigarh Call Girl...
Chandigarh Escorts Service 📞8868886958📞 Just📲 Call Nihal Chandigarh Call Girl...
Sheetaleventcompany
 
0183760ssssssssssssssssssssssssssss00101011 (27).pdf
0183760ssssssssssssssssssssssssssss00101011 (27).pdf0183760ssssssssssssssssssssssssssss00101011 (27).pdf
0183760ssssssssssssssssssssssssssss00101011 (27).pdf
Renandantas16
 

Dernier (20)

Katrina Personal Brand Project and portfolio 1
Katrina Personal Brand Project and portfolio 1Katrina Personal Brand Project and portfolio 1
Katrina Personal Brand Project and portfolio 1
 
Monthly Social Media Update April 2024 pptx.pptx
Monthly Social Media Update April 2024 pptx.pptxMonthly Social Media Update April 2024 pptx.pptx
Monthly Social Media Update April 2024 pptx.pptx
 
Call Girls Pune Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Pune Just Call 9907093804 Top Class Call Girl Service AvailableCall Girls Pune Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Pune Just Call 9907093804 Top Class Call Girl Service Available
 
How to Get Started in Social Media for Art League City
How to Get Started in Social Media for Art League CityHow to Get Started in Social Media for Art League City
How to Get Started in Social Media for Art League City
 
Call Girls In Noida 959961⊹3876 Independent Escort Service Noida
Call Girls In Noida 959961⊹3876 Independent Escort Service NoidaCall Girls In Noida 959961⊹3876 Independent Escort Service Noida
Call Girls In Noida 959961⊹3876 Independent Escort Service Noida
 
VVVIP Call Girls In Greater Kailash ➡️ Delhi ➡️ 9999965857 🚀 No Advance 24HRS...
VVVIP Call Girls In Greater Kailash ➡️ Delhi ➡️ 9999965857 🚀 No Advance 24HRS...VVVIP Call Girls In Greater Kailash ➡️ Delhi ➡️ 9999965857 🚀 No Advance 24HRS...
VVVIP Call Girls In Greater Kailash ➡️ Delhi ➡️ 9999965857 🚀 No Advance 24HRS...
 
Call Girls Ludhiana Just Call 98765-12871 Top Class Call Girl Service Available
Call Girls Ludhiana Just Call 98765-12871 Top Class Call Girl Service AvailableCall Girls Ludhiana Just Call 98765-12871 Top Class Call Girl Service Available
Call Girls Ludhiana Just Call 98765-12871 Top Class Call Girl Service Available
 
Ensure the security of your HCL environment by applying the Zero Trust princi...
Ensure the security of your HCL environment by applying the Zero Trust princi...Ensure the security of your HCL environment by applying the Zero Trust princi...
Ensure the security of your HCL environment by applying the Zero Trust princi...
 
MONA 98765-12871 CALL GIRLS IN LUDHIANA LUDHIANA CALL GIRL
MONA 98765-12871 CALL GIRLS IN LUDHIANA LUDHIANA CALL GIRLMONA 98765-12871 CALL GIRLS IN LUDHIANA LUDHIANA CALL GIRL
MONA 98765-12871 CALL GIRLS IN LUDHIANA LUDHIANA CALL GIRL
 
Business Model Canvas (BMC)- A new venture concept
Business Model Canvas (BMC)-  A new venture conceptBusiness Model Canvas (BMC)-  A new venture concept
Business Model Canvas (BMC)- A new venture concept
 
👉Chandigarh Call Girls 👉9878799926👉Just Call👉Chandigarh Call Girl In Chandiga...
👉Chandigarh Call Girls 👉9878799926👉Just Call👉Chandigarh Call Girl In Chandiga...👉Chandigarh Call Girls 👉9878799926👉Just Call👉Chandigarh Call Girl In Chandiga...
👉Chandigarh Call Girls 👉9878799926👉Just Call👉Chandigarh Call Girl In Chandiga...
 
It will be International Nurses' Day on 12 May
It will be International Nurses' Day on 12 MayIt will be International Nurses' Day on 12 May
It will be International Nurses' Day on 12 May
 
Call Girls In Panjim North Goa 9971646499 Genuine Service
Call Girls In Panjim North Goa 9971646499 Genuine ServiceCall Girls In Panjim North Goa 9971646499 Genuine Service
Call Girls In Panjim North Goa 9971646499 Genuine Service
 
Call Girls From Pari Chowk Greater Noida ❤️8448577510 ⊹Best Escorts Service I...
Call Girls From Pari Chowk Greater Noida ❤️8448577510 ⊹Best Escorts Service I...Call Girls From Pari Chowk Greater Noida ❤️8448577510 ⊹Best Escorts Service I...
Call Girls From Pari Chowk Greater Noida ❤️8448577510 ⊹Best Escorts Service I...
 
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
 
FULL ENJOY Call Girls In Majnu Ka Tilla, Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Majnu Ka Tilla, Delhi Contact Us 8377877756FULL ENJOY Call Girls In Majnu Ka Tilla, Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Majnu Ka Tilla, Delhi Contact Us 8377877756
 
Mysore Call Girls 8617370543 WhatsApp Number 24x7 Best Services
Mysore Call Girls 8617370543 WhatsApp Number 24x7 Best ServicesMysore Call Girls 8617370543 WhatsApp Number 24x7 Best Services
Mysore Call Girls 8617370543 WhatsApp Number 24x7 Best Services
 
Falcon Invoice Discounting platform in india
Falcon Invoice Discounting platform in indiaFalcon Invoice Discounting platform in india
Falcon Invoice Discounting platform in india
 
Chandigarh Escorts Service 📞8868886958📞 Just📲 Call Nihal Chandigarh Call Girl...
Chandigarh Escorts Service 📞8868886958📞 Just📲 Call Nihal Chandigarh Call Girl...Chandigarh Escorts Service 📞8868886958📞 Just📲 Call Nihal Chandigarh Call Girl...
Chandigarh Escorts Service 📞8868886958📞 Just📲 Call Nihal Chandigarh Call Girl...
 
0183760ssssssssssssssssssssssssssss00101011 (27).pdf
0183760ssssssssssssssssssssssssssss00101011 (27).pdf0183760ssssssssssssssssssssssssssss00101011 (27).pdf
0183760ssssssssssssssssssssssssssss00101011 (27).pdf
 

En vedette

Social Media Marketing Trends 2024 // The Global Indie Insights
Social Media Marketing Trends 2024 // The Global Indie InsightsSocial Media Marketing Trends 2024 // The Global Indie Insights
Social Media Marketing Trends 2024 // The Global Indie Insights
Kurio // The Social Media Age(ncy)
 
Good Stuff Happens in 1:1 Meetings: Why you need them and how to do them well
Good Stuff Happens in 1:1 Meetings: Why you need them and how to do them wellGood Stuff Happens in 1:1 Meetings: Why you need them and how to do them well
Good Stuff Happens in 1:1 Meetings: Why you need them and how to do them well
Saba Software
 
Introduction to C Programming Language
Introduction to C Programming LanguageIntroduction to C Programming Language
Introduction to C Programming Language
Simplilearn
 

En vedette (20)

How to Prepare For a Successful Job Search for 2024
How to Prepare For a Successful Job Search for 2024How to Prepare For a Successful Job Search for 2024
How to Prepare For a Successful Job Search for 2024
 
Social Media Marketing Trends 2024 // The Global Indie Insights
Social Media Marketing Trends 2024 // The Global Indie InsightsSocial Media Marketing Trends 2024 // The Global Indie Insights
Social Media Marketing Trends 2024 // The Global Indie Insights
 
Trends In Paid Search: Navigating The Digital Landscape In 2024
Trends In Paid Search: Navigating The Digital Landscape In 2024Trends In Paid Search: Navigating The Digital Landscape In 2024
Trends In Paid Search: Navigating The Digital Landscape In 2024
 
5 Public speaking tips from TED - Visualized summary
5 Public speaking tips from TED - Visualized summary5 Public speaking tips from TED - Visualized summary
5 Public speaking tips from TED - Visualized summary
 
ChatGPT and the Future of Work - Clark Boyd
ChatGPT and the Future of Work - Clark Boyd ChatGPT and the Future of Work - Clark Boyd
ChatGPT and the Future of Work - Clark Boyd
 
Getting into the tech field. what next
Getting into the tech field. what next Getting into the tech field. what next
Getting into the tech field. what next
 
Google's Just Not That Into You: Understanding Core Updates & Search Intent
Google's Just Not That Into You: Understanding Core Updates & Search IntentGoogle's Just Not That Into You: Understanding Core Updates & Search Intent
Google's Just Not That Into You: Understanding Core Updates & Search Intent
 
How to have difficult conversations
How to have difficult conversations How to have difficult conversations
How to have difficult conversations
 
Introduction to Data Science
Introduction to Data ScienceIntroduction to Data Science
Introduction to Data Science
 
Time Management & Productivity - Best Practices
Time Management & Productivity -  Best PracticesTime Management & Productivity -  Best Practices
Time Management & Productivity - Best Practices
 
The six step guide to practical project management
The six step guide to practical project managementThe six step guide to practical project management
The six step guide to practical project management
 
Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...
Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...
Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...
 
Unlocking the Power of ChatGPT and AI in Testing - A Real-World Look, present...
Unlocking the Power of ChatGPT and AI in Testing - A Real-World Look, present...Unlocking the Power of ChatGPT and AI in Testing - A Real-World Look, present...
Unlocking the Power of ChatGPT and AI in Testing - A Real-World Look, present...
 
12 Ways to Increase Your Influence at Work
12 Ways to Increase Your Influence at Work12 Ways to Increase Your Influence at Work
12 Ways to Increase Your Influence at Work
 
ChatGPT webinar slides
ChatGPT webinar slidesChatGPT webinar slides
ChatGPT webinar slides
 
More than Just Lines on a Map: Best Practices for U.S Bike Routes
More than Just Lines on a Map: Best Practices for U.S Bike RoutesMore than Just Lines on a Map: Best Practices for U.S Bike Routes
More than Just Lines on a Map: Best Practices for U.S Bike Routes
 
Ride the Storm: Navigating Through Unstable Periods / Katerina Rudko (Belka G...
Ride the Storm: Navigating Through Unstable Periods / Katerina Rudko (Belka G...Ride the Storm: Navigating Through Unstable Periods / Katerina Rudko (Belka G...
Ride the Storm: Navigating Through Unstable Periods / Katerina Rudko (Belka G...
 
Barbie - Brand Strategy Presentation
Barbie - Brand Strategy PresentationBarbie - Brand Strategy Presentation
Barbie - Brand Strategy Presentation
 
Good Stuff Happens in 1:1 Meetings: Why you need them and how to do them well
Good Stuff Happens in 1:1 Meetings: Why you need them and how to do them wellGood Stuff Happens in 1:1 Meetings: Why you need them and how to do them well
Good Stuff Happens in 1:1 Meetings: Why you need them and how to do them well
 
Introduction to C Programming Language
Introduction to C Programming LanguageIntroduction to C Programming Language
Introduction to C Programming Language
 

Blockmed-v1.18

  • 1. BlockMed: Open Protocol for Medical IoT Wayne Chung, Andrew Lee Dec. 8th, 2018 Abstrac​t We present an open protocol for medical IoT devices that connects payer, provider and patients on a public Ethereum blockchain that allows for trustless, secure smart contract execution. Our protocol allows for BlockMed token, BMD to serve as a utility token which allows 3 core innovations: 1) Distributed encryption/decryption of HIPAA protected medical IoT patient data for prevention of hacking and ransomware type attacks in remote monitoring settings. 2) Augmented deep learning and blockchain platform to allow faster, more efficient development of crowdsourced AI algorithms. 3) Machine to machine payment of smart contract parameters from medical IoT data based on individual and community standards. 1
  • 2. Contents I. Introduction…………………………………………………………… 3 II. Current State ………………………………………………………….. 4 III. Enhanced Cybersecurity ………………………………………………. 5 IV. Crowdsourced AI Algorithms ………………………………………….14 V. Machine to Machine Payments ……………………………………….. 20 VI. Token Economics & Governance……………………………………... 23 VII. Timeline ………………………………………………………………. 24 VIII. Summary ……………………………………………………………… 26 IX. Acknowledgements …………………………………………………… 26 X. References …………………………………………………………….. 27 2
  • 3. I. ​Introduction The rapid proliferation of medical IoT devices recording patient data remotely outside the hospital and clinic by connecting to the mobile-cloud is creating increasing cybersecurity vulnerabilities for the health IT infrastructure. Provider based EHRs are understandably concerned and reluctant to open their APIs to the myriad of remote monitoring devices connecting to the mobile-cloud for fear of hacking and ransomware type attacks. Enhanced cybersecurity of existing health IT and EHRs is top of mind for CIO’s of all major health organizations. BlockMed proposes an open protocol which allows remote medical IoT data to be HIPAA compliant on and off the blockchain so there is record of sensor data verified with the patient, provider and payer. Username-password compromises, as seen in centralized databases allowing current ransomware attacks, will be more difficult on the distributed platform proposed here. The exponential explosion of medical IoT data is becoming a double-edge sword regarding utility and value created from such data. Providers are increasingly overwhelmed from alarm fatigue and false positives which generate unnecessary calls, visits, and further diagnostics, potentially creating more legal liability. The fundamental value of this data resides in the actionable algorithms created to predict and inform therapeutic action. Creating such actionable algorithms depends heavily on the quantity and quality of the data, as compared to the entire body of data from the individual and community. Remote medical IoT data analysis are currently limited to the number of patients which can generate that data and the team members within an individual company to create actionable algorithms from that data. BlockMed proposes a novel open platform whereby such remotely generated medical IoT data can be crowdsourced out to the BMD community to then develop superior algorithms, that overcomes standard overfitting bias based on static historical data as opposed to new continuously generated live data. This deep learning blockchain platform will 3
  • 4. allow faster, more efficient development of crowdsourced AI algorithms, that can then result in machine to machine payment of smart contract value from medical IoT data based on real-time live data as opposed to traditional population health models currently in existence. II. ​Current State Blockchain technology applied to medical IoT is still in its infancy. Decentralized distributed ledger technology is ideal for immutable records of crowdsourced algorithm performance and machine to machine payments where secure identification and verification of algorithm results is crucial for widespread adoption. Many other proposed projects identify the EHR and existing legacy IT systems as starting point for a blockchain based replacement. This rapidly growing eco-system can accommodate many approaches, but in our opinion the medical IoT space is relatively new like blockchain--therefore a prime opportunity to rapidly establish a widely adopted and scalable use case as there is no current EHR or IT legacy system to overcome and replace. Other proposed projects attempt to combine predictive algorithms with blockchain for better disease prediction on existing data sets from current hospital or clinic-based systems. Such projects have issues of ownership of data, as the existing legacy hospital and clinic systems are reluctant to enter into profitable agreements to share this data for the proposed types of blockchain projects. Without this legacy data in actionable formats from current hospital or clinic-based IT systems, blockchain development of predictive algorithms in these types of projects may be much slower in getting massive data to scale and yielding superior actionable results. Supply chain management of hospital and medical device supplies are other proposed blockchain projects. Given the complicated logistics involved with large legacy systems, this area is better suited for large existing IT vendors such as IBM, Microsoft, AWS to offer blockchain services 4
  • 5. for large companies to deploy--as IBM has already done with the cargo shipping industry. Simply replicating entire aspects of legacy health IT infrastructures with blockchain is unlikely to gain rapid adoption and scale if the complexity of legacy systems and lack of data access continues--as will likely be the case while the legacy incumbents slowly adopt blockchain on their own. Areas which fundamentally lends itself to open protocol development have the best chance for rapid network adoption and true scaling potential for existing healthcare players. BlockMed is proposing an open protocol which allows medical IoT data largely untethered to current legacy IT systems, to be rapidly adopted by payers, providers, and patients in a BMD token eco-system with core innovations in enhanced cybersecurity, crowdsourced AI algorithms and machine to machine payment. III. ​Enhanced Cybersecurity BlockMed will utilize a patented wearable multi-functional sensor​[1]​ to initially provide data for its proposed open protocol for enhanced cybersecurity of medical IoT. This proprietary sensor has been developed for over 5 years by Wireless Medical for remote monitoring of cardiac disease. It produces 9 real-time data streams from a credit card size sensor worn on chest that sends data via Bluetooth to an iOS app on iPhone, which streams the data to AWS cloud server, where it is stored and sent back to iOS app for display of data to physician. For HIPAA compliance, the data streamed from sensor via Bluetooth will be encrypted with AES-128 protocols. The data streamed from iOS app to web app will be protected by SSL encryption, while the data stored in AWS cloud server will have dedicated hardware provided for encrypted data at rest. For sensor’s 9 real-time data streams, no such encryption is necessary as long as there is no pairing with the HIPAA compliant patient identifiers. The raw sensor data will be stored on public distributed file storage protocols such as IPFS​[2]​ via BlockMed open 5
  • 6. source API​[3]​ (See Figure 1). This not only enables other technology and framework such as IoT Fog/Edge computing framework proposed by Cisco​[4]​ to take advantage of IPFS that has better performance than others such as Amazon S3​[5]​ , but also shows better performance gain with IPFS​[6]​ . Figure 1 The proposed open protocol will utilize a public Ethereum​[7]​ blockchain for the smart contract execution, while the HIPAA compliant patient identifier data will be stored off-chain such as an AWS encrypted databases as necessary for proper onboarding of medical sensor usage. The private keys of public Ethereum blockchain can be kept on mobile app of patient. In case of loss of mobile app/smart phone, this unique key can have appropriate recovery or replacement mechanisms provided by BlockMed​[8]​ . Furthermore, the API kit can be extended to support data encryption on top of IPFS as an extra layer of security to keep data encrypted and give users full control of their data. In addition to user data, the open protocol also provides a channel to compute Private Contracts, in other word, confidential code that can be service through off-chain service provider that provides TEE (Intel SGX) backed compute-nodes​[9][10]​ . In Oasis Lab’s Ekiden​[11]​ protocol, clients send inputs to confidentiality-preserving smart contracts, which are executed within a TEE at any compute node. The blockchain stores encrypted contract state. 6
  • 7. This provides opportunity for regulated industries to utilize the open protocol and compute public data from IPFS along with private data off-chain with proprietary code. The owner of the data who holds the Private Key can authorize different service providers with the proper key to decrypt the data and still track of each service provider on how the authorization are distributed and applied. Other services such as Enigma​[12]​ can also take advantages to corporate data privacy off-chain​[13]​ by exchanging the authorized secret key access from user via Ethereum smart contract while the data are still kept on the public IPFS for further usage. Enigma also opens the door to multi-party computation, MPC​[14][15]​ as well. This also leaves a trace of who accessed the data on the public chain as an evidence and audit. (See Figure 2) Figure 2 The distributed ledger technology of public Ethereum blockchain provides enhanced security with immutable records of transactions on nodes. As such, it is very difficult to hack in traditional ways seen in centralized databases such as AWS, often through username-password compromises seen frequently in ransomware attacks on hospital systems. BlockMed’s proposed open protocol cooperating with private off-chain, AWS encrypted database for patient, physician ID, and public IPFS file storage for raw sensor data serves as redundancy for both EHRs’ and Payers’ centralized databases. (See Figure 3) 7
  • 8. Figure 3 In cases of ransomware attacks on institutions, utilizing BlockMed’s proposed open protocol, the distributed and decentralized private keys stored on patients’ mobile apps provides an additional level of security for the medical IoT data. As a result, there may be little incentive to pay ransomware demands as the distributed IPFS file storage data can be easily recovered and duplicated onto legacy systems. (See Figure 4) Attack Mitigation The following illustrate how a potential attack scenario by Ransomware or compromising the server managed by central authority for business usage. 8
  • 9. Figure 4 - Attack Scenario with Ransomware According to Figure 4, immutable records and data are already anonymized and de-identified on IPFS with the list of pointers hosted by one or more Open Source Web Server​[16]​ (via Caching and Indexing). This means by gaining access to the server or the Open Source Web Server will limit the chance to correlate and expose patient’s identities based on the randomized UUID. This Open Source Web Server can be replicated and distributed by anyone anywhere for high availability, reliability. and performance​[17]​ . The authority can decide how many of the caching servers they would like to host internally to reach a quorum for reliability in their own environment, or simply become part of the public Internet that interacts with other alike servers (however, this is likely to happen due to each authority will like to keep track of a minimal subset of the global records on IPFS, especially from their own patients). As a result, the victim authority will be able to perform the following to recover from such attack. (See Figure 5) 1. Setup a new secured server with latest security patch, etc. 2. Migrate the IP address from the old server to the new one 9
  • 10. a. Existing sensors will retry to register on this ‘new’ server, and afterward, continue to send data to it. This provides the server a list of existing devices (aka previous registered patients). b. In addition to the new incoming data after the service availability is restored, without the existing sensors sending data, the ‘new’ server can coordinate with the Open Source Web Server to pre-load all the historical index to fetch the data from IPFS. This will allow the ‘new’ server to bootstrap the lost historical data, meanwhile, still receive new (incremental data) from the existing sensors. 3. After a certain period of time, the ‘new’ server will gain its original view prior to the attack, and hence, can mitigate the attack (e.g. ransomware). Figure 5 - Recovery Scenario to Restore Data 10
  • 11. BlockMed’s proposed open protocol allows for developers worldwide to contribute to the sensors’ raw data storage through the BlockMed token, BMD. Wireless Medical’s proprietary sensor’s 9 real-time data streams accumulating 24/7 daily quickly adds up to large amounts of data. Blockchain technology as applied to healthcare IT security is still in its infancy. Two main issues BlockMed will address along with entire public Ethereum blockchain community include: 1) tradeoffs between maintaining security while increasing transaction speeds to handle increasingly large amounts of raw data on the public Ethereum blockchain and IPFS file storage network, (See Figure 6) Figure 6 - Tradeoff Between Ethereum and IPFS Storage 2) having useful off-chain API’s to interact with public Ethereum blockchain and IPFS file storage networks in order to handle all the various medical IoT devices, EHR’s, Insurance Payers, App/Browser eco-systems etc. that will be necessary for future scaling during mainstream adoption phase. BlockMed’s internal team and open community developers will use BMD token incentives to contribute data from Wireless Medical’s multi-functional sensor as well as their own off the shelf 11
  • 12. sensor data to the distributed IPFS file storage platform. Consumer devices with medical applications such as Apple Health Kit API, Apple Watch, Fitbit and other app systems such as Android, and other web browser-based systems can also be tested and developed on BMD open platform by developers. This growth will be powered by worldwide community of developers incentivized through BMD token to improve functionality and security of open platform for medical IoT devices. (Figure 7) Figure 7 - Multiple Wearable Devices Integrated with Ethereum and IPFS This platform will initially interface with the existing Swift iOS app, Ruby-Python server and MongoDB database residing on AWS. (See Figure 8) 12
  • 13. Figure 8 - Initial Implementation with IPFS and BMD Smart Contract As improvements are made to the distributed IPFS file storage platform handling the open communities’ sensor data, consensus mechanisms will be utilized through BMD smart contracts to incentivize changes that achieve stated goals of increasing speed and efficiency in handling raw sensor data on distributed IPFS as well as usefulness in interacting with appropriate API’s for off chain EHR’s such as Epic which is used in Stanford, UCSF hospitals, where initial pilots will be conducted, and API’s for off chain Insurance Payers such as United Health, Humana, or others which may enter into agreements during pilot studies. In Stanford, UCSF pilot studies, the patient and physician ultimately should not notice any difference in testing the Wireless Medical multi-functional sensor and mobile app with regards to the data streams collected, between non-blockchain and BMD blockchain versions for enhanced cybersecurity. (See Figure 9) 13
  • 14. Figure 9 - Interaction with Ethereum and IPFS are Transparent After successful pilot studies, emphasis will be on API integration and bringing on commercial partners onto BlockMed open platform with other hospital systems, EHRs, Payers, and other medical IoT devices. IV. ​Crowdsourced AI Algorithms BlockMed’s 2​nd​ core innovation is that Wireless Medical’s multi-functional proprietary sensor algorithms can be crowdsourced real-time to open community through use of BMD tokens to produce Dutch​[18]​ style multi-unit auctions that result in superior machine learning algorithms on each of the 9 sensor data streams, as well as a combined composite score. This technique can overcome issues​[19]​ of overfitting algorithms​[20]​ to static historical test data that are poor performing on newer real-time data streams. This is critical in the medical IoT space, not only for Wireless Medical but other medical IoT devices, which all constantly produce new 24/7 real-time sensor data under wide variety of variable environmental conditions. The 9 data streams produced by Wireless Medical’s sensor include: 1) acoustic heart sounds 2) skin impedance 3) EKG 4) heart rate 5) heart rate variability 6) respiratory rate 7) temperature 8) sleep angle 9) steps walked. Each of these data streams represent physiological states in 14
  • 15. patient wearing sensor, but exist in noisy remote, non-clinical environment where movement and other factors can introduce artifacts into data streams. (See Figure 10) Figure 10 - De-identified Input Stream from IPFS to AI Frameworks In proposed open protocol, the sensor’s raw data streams, de-identified on distributed IPFS file storage network, can be made available to community of BMD token holders and developers, who can utilize various machine learning and deep learning algorithms frameworks such as TensorFlow, PyTorch, CNTK, and MXNet ​to more efficiently refine accuracy not only on static, old historical data, but on constantly generated new live data. (See Figure 11) Figure 11 - General AI Framework Tapping into BMD Protocol 15
  • 16. A multi-unit Dutch auction can be utilized so algorithms performing better on new data get incentivized through BMD tokens issued to these developers in regularly scheduled contests​[21]​ . Developers can also use their own off the shelf consumer devices for commodity data streams such as HR, HRV, RR, Temp, Sleep Angle and Steps to submit algorithms which can be tested against other consumer devices as well as Wireless Medical’s proprietary multi-functional sensor. Complex, unique data streams such as acoustic heart sounds, skin impedance and EKG’s can only be gathered from Wireless Medical’s proprietary sensor. Initially this data will be gathered and tested on healthy developers from BlockMed’s internal team with access to proprietary sensor, but later on sick patients in the UCSF, Stanford clinical studies. These complex, unique data streams can also be crowdsourced and incentivized through a multi-unit Dutch auction to BMD developer community just like the commodity data streams. The Dutch auction’s goal will be to produce algorithm models which accurately predict behavior of both individual data streams and combined sensor score as related to true physiological state of patient. For example, raw HR data gathered from Wireless Medical’s multi-functional sensor can be used to create predictive algorithm models, then tested on Apple Watch’s raw HR data and vice versa. These models will need to account for occasional falsely high HR’s due to wearable environment as compared to clinically labelled HR as gold standards. Best models submitted by BMD token community of developers can be forward tested on wide variety of other devices new raw HR data for a continuously evolving winning model. Mechanics of reward amounts, schedule and winners can be governed by existing best practices and modified by BMD community through consensus. This Dutch auction can be applied to all 9 individual data streams and combined composite score produced by Wireless Medical’s multi-functional sensor. (See Figure 12) 16
  • 17. Figure 12 - Derived and Published Algorithm from Crowdsourced AI gets Rewarded The BMD crowdsourced machine learning algorithms can also be applied to the mobile data generated through future voice activated app platform along with camera, video data upon app interaction by patient. True artificial intelligence through applying deep learning natural language processing and computer vision techniques to mobile audio​[22]​ and video​[23][24]​ data streams along with machine generated sensor data and be compared to human labelled clinical data in clinical studies over a longitudinal time frame of 7-day hospital stay and 30-day discharge period at home. A combined composite mobile and sensor score of all 9 data streams equally weighted can be compared to equivalent composite clinical score of 9 physician reported parameters and compared for correlation and predictive power in the hospital and later in patient’s home environment. Again, by crowdsourcing to BMD token developer community, the efficient optimization of superior algorithms—a world’s first true deep learning platform on the blockchain can be created not only for Wireless Medical’s multifunctional sensor, but ultimately 17
  • 18. for other medical IoT devices generating massive amounts of 24/7 physiological data real time​[25]​ . Figure 13 - Data Published to IPFS and BMD Smart Contract The BMD token developer community can be incentivized to analyze open source community data such as Beth Israel—MIT EKG database for arrhythmias, or TI’s database for HR detection algorithms and efficiently optimize algorithms on both individual and community level. Ultimately the goal will be to generate supervised, semi-supervised and unsupervised deep learning algorithms​[26][27]​ that trains on labelled and unlabeled open source community data and individual medical IoT data to perform accurately on newly generated patient data across wide variety of settings, as that will prove a most powerful and valuable use case. Unlike previous and existing legacy systems, this open protocol allows the BMD token developer community to capture much of this value creation in the deep learning algorithms of newly generated mobile and sensor data as shown to be useful for the patient and physician. The de-identified raw sensor data on public distributed IPFS file storage network will always be kept separate from patient, physician identifiers kept on off-chain, encrypted AWS. ​One 18
  • 19. possibility is to run Apache Spark​[28]​ on Amazon EMR, and use Terraform​[29]​ (by HashiCorp) to build a Spark and Apache Zeppelin​[30]​ cluster on Amazon EMR which is HIPAA compliant​[31]​ . This solution encrypts all data at rest and in-flight, logs all user activities, as well as satisfies many other aspects of a HIPAA compliant environment. The use of Terraform provides a high degree of management of Cluster Configuration, Data Accessibility, Scalability, Security, and Availability. (See Figure 14) Figure 14 - Running Big Data Applications in HIPAA Environment As such HIPAA compliance can be met, while allowing open BMD developer community to rapidly iterate, and test mobile and sensor data with real world clinical inputs to create a deep learning blockchain platform which can be applied not only to affect various cardiac conditions such as heart failure, cardiac arrhythmias, post heart surgery management but also other non-cardiac entities such as pulmonary diseases etc. 19
  • 20. V. ​Machine to Machine Payments BlockMed’s 3​rd​ core innovation is to allow machine to machine payments on the open Ethereum blockchain platform, such that any set of individual and community parameters can be set in the smart contract as incentives for individual data streams and combined sensor score results. Current medical IoT devices have to go through large, expensive studies to prove positive changes in clinical outcomes worthy of reimbursement by Payers. These population-based studies provide a major hurdle to widespread medical IoT adoption as the desired outcomes measure results in lengthy binary decisions to reimburse or not based on entire population tested when certain patients may benefit at individual level. (See ​Figure 15) Figure 15 – Current Medical IoT Ecosystem The BMD open platform will allow Payer to incentivize each sensor data stream and combined score to standards on individual real-time level. For example, HR can be tracked to certain normal ranges, and deviations from that community standard can be verified on public Ethereum blockchain platform. Patients’ that have minimal deviations can then have de-identified BMD token transfers on public Ethereum blockchain that connects to Payer’s off-chain encrypted account of that individual with resulting co-pays, deductibles etc. available to adjust accordingly. Such value transfers with BMD tokens real-time from Payer to device at individual level allows for faster commercial adoption of medical IoT devices as Payers do not necessarily need to wait until binary outcomes data from large commercial adoption studies when deciding on reimbursement in remote monitoring setting. Instead, medical IoT devices can be incentivized to 20
  • 21. join open platform with Payers setting individual standards for data streams regarding BMD token value transfers. (See ​Figure 16) Figure 16 - BlockMed Token Ecosystem Positive reinforcement for clinical compliance as evidenced by medical IoT data that adheres to community standards can now be transferred through BMD tokens real-time, machine to machine. Payers that allow select patient community data to be de-identified and made available to open platform can benefit from crowdsourced modelling of most efficient payment incentives at individual level, similar to Dutch auction model used for deep learning algorithm development regarding accuracy of individual and composite data streams. Again, the BMD platform allows developers who create the most accurate algorithms for Payers to transfer token value can capture most of the value creation, as there is much potential cost reduction in entire healthcare system regarding prevention and management of chronic disease in remote monitoring setting. (See ​Figure 17) 21
  • 22. Figure 17 - Network Virality Effect and BMD Circulation The medical IoT space is ideal for machine to machine payments as typical hospital re-admissions can cost Payer $10-15K, and happen 20-25% of time on all major admission categories such as heart failure, pneumonia, heart surgeries etc.​[32]​ . In addition, patient deductibles are increasingly high, such as $3-10K on all types of Payer plans. Crowdsourced deep learning models which accurately predict individual data streams at individual level, then tie combined composite score directly to Payer accounts of deductibles and premiums, all on off-chain, encrypted AWS and public Ethereum blockchain platforms offer unique opportunity to transform fundamental payment models currently in place primarily for inpatient and clinic visits. (See ​Figure 18) 22
  • 23. Figure 18 - Enterprise and Secure Computation in BMD Network Reimbursement for medical IoT remote monitoring is still very much in its infancy due to lack of quantifiable data regarding effectiveness of various low-tech disease management approaches as well as digital therapeutic, medical IoT offerings. BMD’s open platform allows for quantified, results orientated, machine to machine micro-payments at individual level, based on medical IoT real-time data streams. VI. ​Token Economics & Governance BlockMed will issue its BMD tokens through pre-sale and later it in public sale on rolling basis until funds deemed sufficient for network development. A trusted public wallet managing the escrow account with the initial 1B tokens such as CoinBase with secure offline cold storage 23
  • 24. capabilities will be used to safely handle the exchange of Ethereum, Bitcoin, USD, Euro into BMD tokens. Appropriate identification provided by third party vendors will be requested so Know Your Customer (KYC), Anti-Money Laundering (AML) and all other relevant regulations are complied with. A multi-sig. contract whose keys are held by trusted individuals within BlockMed will be used to handle all funds. 1 billion total BMD tokens will be issued at initial exchange rate of 1USD per BMD token. Allocations include: 20% to founding team and advisors, 30% to BMD Foundation, 50% to public. (See ​Figure 19) Figure 19 – BlockMed, BMD ERC20 Smart Contract Execution BlockMed’s initial governance will be guided by founding team, board of directors to set direction of strategy, including key internal hires and outside engineering design/ development houses to execute out private and public blockchain platforms in preparation for UCSF, Stanford clinical pilot and later commercial network Payer partners. To promote a common standard among the open protocol proposed by BlockMed, decentralized consensus may be utilized with open governance and voting among the BMD token holders, guided by BlockMed’s founding 24
  • 25. team. Later a true decentralized autonomous organization (DAO) governance structure may be considered once the open protocol has reached sufficient community adoption. Figure 20 - Token Allocation VII.​ ​Timeline Jan.--Aug. 2018—Whitepaper Research May--Dec. 2108—Smart Contract Prototype Jan.--Dec. 2019—Rolling Raise BMD Tokens Jan.--Dec. 2019—BlockMed.AI Platform Development July--Dec. 2019—UCSF, Stanford Pilot Studies Jan.--Dec. 2020—Payer / Pharma Licensing 25
  • 26. VIII. ​Summary Remote medical IoT data has the potential to fundamentally transform the healthcare landscape with regards to chronic care disease management, keeping patients out of the hospital and allowing more cost savings to entire system through early prediction and therapeutics. Unfortunately, the current system depends on large scale trials and pilots with payers and providers to prove cost savings on community populations before large scale deployment and reimbursement. BlockMed proposes a novel, secure, deep learning blockchain platform, based on the BMD token, whereby large population-based community trials are not a necessary precursor to individual transfer of value real time based on medical IoT data. Based on public Ethereum smart contract parameters, remote medical IoT data that meets certain algorithmic criteria can result in real-time machine to machine payments through BMD token. Such a protocol has potential to transform transmission of value throughout the entire payer, provider, patient ecosystem for medical IoT, and truly incentivize improved health outcomes and lower costs among all affected parties. IX. Acknowledgements We would like to thank our mentors, advisors, and friends who have provided invaluable advice on BlockMed. In particular, for educating, reviewing and providing feedback on this work in specific and blockchain-cryptocurrency world in general. 26
  • 27. X. References [1] WMMIP101US; U.S. Application Serial No. 14/050,356; Issue Notification 08-21-18 [2] ​https://github.com/ipfs/papers/raw/master/ipfs-cap2pfs/ipfs-p2p-file-system.pdf​ and J. Benet, “IPFS - Content Addressed, Versioned, P2P File System,” Protocol Labs, Inc., Tech. Rep., 2014. [3] ​https://github.com/blcksync/bc-ipfs [4] F. Bonomi, R. Milito, J. Zhu, and S. Addepalli, “Fog Computing and Its Role in the Internet of Things,” in ​Proceedings of the First Edition of the MCC Workshop on Mobile Cloud Computing​, ser. MCC ’12, 2012, pp. 13–16. ​https://github.com/ethereum/wiki/wiki/White-Paper [5] B. Confais, A. Lebre, and B. Parrein, “Performance Analysis of Object Store Systems in a Fog/Edge Computing Infrastructures,” in ​IEEE CloudCom​, Luxembourg, Luxembourg, Dec. 2016. [6] ​https://hal.archives-ouvertes.fr/hal-01483702/document [7] ​https://github.com/ethereum/wiki/wiki/White-Paper [8] ​https://github.com/bitcoin/bips/blob/master/bip-0044.mediawiki [9] Fisch, B., Vinayagamurthy, D., Boneh, D., and Gorbunov, S. Iron: functional encryption using Intel SGX. In ​Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security​ (2017), ACM, pp. 765–782 [10] Nayak, K., Fletcher, C., Ren, L., Chandran, N., Lokam, S., Shi, E., and Goyal, V. Hop: Hardware makes obfuscation practical. In ​24th Annual Network and Distributed System Security Symposium,​ NDSS (2017) [11] Ekiden: A Platform for Confidentiality-Preserving, Trustworthy, and Performant Smart Contract Execution ​https://arxiv.org/abs/1804.05141 [12] ​https://enigma.co/protocol/SoftwareArchitecture.html [13] Bogdanov, Dan, Sven Laur, and Jan Willemson. ”Sharemind: A framework for fast privacy- preserving computations.” ​Computer Security-ESORICS​ 2008. Springer Berlin Heidelberg, 2008. 192-206 [14] Yao, Andrew C. ”Protocols for secure computations.” 2013 IEEE ​54th Annual Symposium on Foundations of Computer Science​. IEEE, 1982 27
  • 28. [15] Lindell, Y., and Pinkas, B. Secure multiparty computation for privacy- preserving data mining. ​Journal of Privacy and Confidentiality​ 1, 1 (2009), 5 [16] ​https://en.wikipedia.org/wiki/Comparison_of_web_server_software [17] Consistency of Data replication protocols in database systems: a review - ​International Journal on Information Theory (IJIT),​Vol.3, No.4, October 2014 [18] ​Vickrey, W. (1961). Counter speculation, auctions, and competitive sealed tenders. The Journal of Finance​, 16(1), 8–37. A pathbreaking paper that introduced second price auctions and performed new analysis of first price [19] Hardt. Competing in a data science contest without reading the data. Http://blog.mrtz.org/2015/03/09/competition.html​​ [20]​​ Dwork, Feldman, Hardt, Pitassi, Reingold, Roth. Generalization in Adaptive Data Analysis and Holdout Reuse. http://papers.nips.cc/paper/5993-generalization-in-adaptive-data-analysis-and-holdout-reuse.pdf​. [21] Numerai - ​https://numer.ai/static/media/whitepaper.29bf5a91.pdf [22]​​ Chan, W.; Jaitly, N.; Le, Q.; and Vinyals, O. 2016. Listen, attend and spell: A neural network for large vocabulary conversational speech recognition. In ​​Acoustics, Speech and Signal Processing (ICASSP), 2016 IEEE Inter- national Conference on​​, 4960–4964. IEEE [23] Esteva, A.; Kuprel, B.; Novoa, R. A.; Ko, J.; Swetter, S. M.; Blau, H. M.; and Thrun, S. 2017. Dermatologist-level classification of skin cancer with deep neural networks. ​​Nature 542(7639):115–118 [24] Gulshan, V.; Peng, L.; Coram, M.; Stumpe, M. C.; Wu, D.; Narayanaswamy, A.; Venugopalan, S.; Widner, K.; Madams, T.; Cuadros, J.; et al. 2016. Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. Jama ​316(22):2402–2410 [25] Tison, G. H.; Singh, A. C.; Ohashi, D. A.; Hsieh, J. T.; Ballinger, B. M.; Olgin, J. E.; Marcus, G. M.; and Pletcher, M. J. 2017b. Abstract 21042: Cardio- vascular risk stratification using off-the-shelf wearables and a multi-task deep learning algorithm. ​​Circulation ​​136(Suppl 1):A21042–A21042 28
  • 29. [26] Dai, A. M., and Le, Q. V. 2015. Semi- supervised sequence learning. In ​​Advances in Neural Information Processing Systems​​, 3079–3087 [27] LeCun, Y.; Bengio, Y.; and Hinton, G. 2015. Deep learning. ​​Nature ​521(7553):436–444 [28] ​https://spark.apache.org/ [29] ​https://www.terraform.io/ [30] ​https://zeppelin.apache.org/ [31] ​https://aws.amazon.com/compliance/hipaa-eligible-services-reference/ [32] 2010 ​International Cochrane Collaboration​ review of 25 RCT trials: 9,500 CHF patients, remote monitoring services resulted in 33% drop of mortality. In contrast, nurse monitoring of heart failure patients by telephone was not shown to reduce mortality, although it was linked to lower readmission rates. Both approaches cut overall health costs. 29