Rise of the Intelligent Internet of Things

Transcription

1 Rise of the Intelligent Internet of Things Opportunities and Challenges of Connected Patients, Environments and Machines Mark Wolff, M.S., Ph.D. Advisory Industry Consultant Chief Health Analytics Strategist Global IoT Division SAS Institute

2 Connected Patients, Devices and Environments While the individual man is an insoluble puzzle, in the aggregate he becomes a mathematical certainty Sherlock Holmes Sign of the Four 1890 CONNECTIVITY

3 in the Beginning Reasoning Foundations of Medical Diagnosis The purpose of this article is to analyze the complicated reasoning processes inherent in medical diagnosis. The importance of this problem has received recent emphasis by the increasing interest in the use of electronic computers as an aid to medical diagnostic processes. Before computers can be used effectively for such purposes, however, we need to know more about how the physician makes a medical diagnosis. Medicine is a science of uncertainty and an art of probability. Sir William Osler ( ) It is predicted that digital electronic computers will assume an increasingly important role in medicine Future programs, in all probability, will facilitate communication and record-keeping in the hospital, relieve the physician of much routine history-taking, calculate diagnostic probabilities, and indicate those diagnostic and therapeutic procedures most likely to benefit the patient. William R. Best, M.D. (1962) Since the earliest days of computers, health professionals have anticipated the day when machines would assist in the diagnostic process Edward H. Shortliffe, M.D., Ph.D. (1987) In Brazil and India, machines are already starting to do primary care, because there s no labor to do it. They may be better than doctors. Mathematically, they will follow evidence and they re much more likely to be right. Robert Kocher, M.D. (2013) "If models based on patient, tumor and treatment characteristics already out-perform the doctors, then it is unethical to make treatment decisions based solely on the doctors' opinions. Cary Oberije, Ph.D. (2013) 1959 The mathematical techniques that we have discussed and the associated use of computers are intended to be an aid to the physician. This method in no way implies that a computer can take over the physician's duties. Quite the reverse; it implies that the physician's task may become more complicated. The physician may have to learn more; in addition to the knowledge he presently needs, he may also have to know the methods and techniques under consideration in this paper. However, the benefit that we hope may be gained to offset these increased difficulties is the ability to make a more precise diagnosis and a more scientific determination of the treatment plan.

4 Advanced Analytics Healthcare Physicians have a reputation as being early adopters of technology. They were early adopters of the very first telephone exchanges built in the 1870 s. In 1906, The Journal of the American Medical Association (JAMA) published the following statement; To no class is the development of the automobile of more importance than to physicians. How to reach their patients in the quickest, surest, easiest and cheapest manner is a practical problem to them.

5 Data, Data, Everywhere but where is the Connectivity? Patient

6 MYCIN 1973 Mycin was a program that used an inferential rules based expert system to help physicians identify serious bacterial infections, such as bacteremia and meningitis, and to recommend antibiotics and appropriate dose to treat the specific infection. Research at Stanford showed that Mycin performed better at identifying specific types of infection better than infectious disease experts and it did so with only 600 rules encoded in the system. Mycin was a groundbreaking approach in supporting clinical decision-making. For the most part the challenges were the limitations of the technology, not the idea, its application or its capabilities. KNOWLEDGE BOTTLENECK and CONNECTIVITY

7

8 The Hyper Connected Patient A New Reality

9 Internet of Medical Things IoMT Applications and hardware that can connect to healthcare information technology systems via networking technologies. Connected People, Devices and Environments facilitate the transfer of clinically relevant data over networks Integration of these data within monitoring, alerting and decision support systems will enable dramatic transformation in the development and manufacture of effective and safe medicinal products as well as accelerating the adoption of Value-Based Health Care.

10 Why IoT Will Give Rise to Artificial Intelligence Forbes Magazine Patterns, problems, and correlations will be easier to address. Intelligent automation will make huge strides leading to a revolution in predictive analytics and proactive intervention will be truly possible. Machine learning may start with chatbots, but AI is the true potential of IoT. The processing of this data (and likely the interpretation and learning of it) will happen in the edge-computing realm. This will be fast and uninhibited.

11 IoT Won t Work Without Artificial Intelligence WIRED MAGAZINE IoT will produce a treasure trove of big data data that can Predict accidents and crimes Give doctors real-time insight into information from pacemakers or biochips Enable optimized productivity across industries through predictive maintenance on equipment and machinery, create truly smart homes with connected appliances Provide critical communication between self-driving cars.

12 Deep Learning Why Now? Many developments are now advancing deep learning: Algorithmic improvements have boosted the performance of deep learning methods. New machine learning approaches have improved accuracy of models. New classes of neural networks have been developed that fit well for applications like text translation and image classification. We have a lot more data available to build neural networks with many deep layers, including streaming data from the Internet of Things, textual data from social media, physicians notes and investigative transcripts. Computational advances of distributed cloud computing and graphics processing units have put incredible computing power at our disposal. This level of computing power is necessary to train deep algorithms.

13 Intelligence Definition The ability to acquire and apply knowledge and skills. Latin intelligentia, from intelligere understand Human Intelligence Machine Intelligence

14 Analytics Machine Intelligence Machine Intelligence

15 Intelligence Completely Automated Public Turing Test To Tell Computers and Humans Apart There are two important properties that any system that mediates robust object recognition must have; INVARIANCE - The ability of the system to respond similarly to different views of the same object. SELECTIVITY - Systems components... produce different responses to potentially quite similar objects (such as different faces) even when presented from similar viewpoints. It is straightforward to make detectors that are invariant but not selective or selective but not invariant. The difficulty lies in how to make detectors that are both selective and invariant. pnas.org/content/early/2013/06/21/

16 How to Win at Atari Breakout Machine Leaning and Optimization Maximize Score Minimize Misses

17 Deep Learning It s Time Instead of organizing data to run through predefined equations, deep learning sets up basic parameters about the data and trains the computer to learn on its own by recognizing patterns using many layers of processing.

18 Analytics at Rest Data Data Storage ETL Deploy Alerts - Reports Decisioning Streaming Data Access - Store - Analyze

19 Enrich Store Deploy Streaming Analytics Including Multiphase Analytics Data Data Storage ETL Train Score Deploy Alerts - Reports Decisioning Streaming Data Streaming Model Execution Supervise Score Train

20 Edge Analytics Intelligent IoT The edge includes physical devices or sensors that interact with the world and collect data. They can be very small in terms of their computing power, perhaps only able to collect their data and transmit it over an industrial protocol. They can also be very powerful, with built-in computing and the capability to route data directly over the network. They are the data producers. By 2019, at least 40% of IoT-created data will be stored, processed, analyzed and acted upon close to, or at the Edge. Source: IDC, Worldwide Internet of Things 2017 Predictions, November 2016

21 The Interconnectivity/Autonomy Human Nervous System People, A Model Environments for Intelligent and IoT? Machines Simply collecting data from connected citizens, sensors, systems or products is not enough. To benefit from Cloud the Analytics promise of IoT data, it will be necessary to shift analytics from traditional data centers toward devices on the edge the things. The challenges The arise Internet from the complexity and risks inherent in capturing and analyzing extreme volumes and varieties of the data torrents owing from ever- increasing numbers of things. Sensor/EA Sensor/EA Sensor/EA Sensor/EA Sensor/EA Sensor/EA Intelligent Internet of Things Copyright Copyright SAS Institute SAS Institute Inc. All Inc. rights All rights reserved. reserved.

22 Application of Sensor Based Analysis of Movement for Quantification of Pain, Discomfort and Disequilibrium Mark Wolff, M.S., Ph.D. Advisory Industry Consultant Global IoT Division SAS Institute J. Gregory Massey, DVM, MSA Manager Heath and Life Sciences Global Hosting and US Professional Services SAS Institute Clinical Decision Support Safety Efficacy Compliance Quality of Life Pre/Rehabilitation Patient Flow Resource Allocation Patient Satisfaction

23 Hypothesis Pain and Discomfort can be Quantified with an iphone

24 Sensor Play Data Stream Collection

25 Motion Sensing Hardware Beyond the iphone

26 iphone Data Collection Methods iphone was mounted to the subject s shirt at the upper right chest/shoulder area. All Data were recorded at 100 Hz Each Subject walked at least 20 seconds in a straight line Training data were recorded from a Subject who is Arthritic Data were recorded in Pain Free and Arthritic States Parameters recorded: Accelerometer Gyroscope

27 Data Accelerometer Example

28 Machine Learning and Modeling

29 Machine Learning and Modeling

30 Balance disorder

31 CONTROL Emilia TREATMENT Isabella Identical Twins Blinded Analysis Each twin walked at least 20 seconds with sensor attached, three times The Analyst (Father) did not know which data belonged to which twin One twin has a much more severe case of Enthesitis The enthesitis is the primary site of disease in spondylitis (AS), a type of arthritis in which there is long term inflammation of the joints of the spine. Typically the joints where the spine joins the pelvis are also affected. Occasionally other joints such as the shoulders or hips are involved.

32 Twin Twin Data Data Gyroscope Two distinct clusters of 3 walks each were identified SAMPLE 1 SAMPLE 2 SAMPLE 3 SAMPLE 4 SAMPLE 5 SAMPLE 6 SAMPLE 1 SAMPLE 2 SAMPLE 3 SAMPLE 4 SAMPLE 5 SAMPLE 6

33 Blinded Results CLUSTER 1 CLUSTER 2 Nominal Walk Arthritic Nominal Walk Walk Arthritic Walk Sample 1 59% Sample 141% 98% 2% Sample 2 65% Sample 235% 97% 3% Sample 3 55% Sample 345% 97% 3%

34 Blinded Results The twin with the more serious case of Enthesitis produced equivocal results Nominal Walk CLUSTER 1 Arthritic Walk The Control model was based on an adult arthritic male Symptoms of Enthesitis, deviated from a nominal walk but were distinct from the control arthritic Walk Sample 1 Sample 2 Sample 3 59% 41% 65% 35% 55% 45%

35 Twin Results The twin with a very mild case of Enthesitis was overwhelmingly identified with a nominal walk Nominal Walk CLUSTER 2 Arthritic Walk Our hypothesis is that nominal can be generalized across gender and age. Sample 1 Sample 2 98% 2% 97% 3% Specific Pathologies will need to me modeled in order to generalize Sample 3 97% 3%

36 Detecting Motor Impairment Disequilibrium Hypothesis: Multi-axial sensors can differentiate impairment due to multiple causes. Tetrahydrocannabinol Ethanol

37 Testing Methodology Posture Stand - feet hip width apart Eyes closed Arms outstretched Head tilted back Time = 30 seconds 100 observations/second 9 sensor variables

38 Developing the Algorithm Test subjects Male & female Age range Control Group No alcohol Treatment Group Wine, beer, liquor

39 Results: Movement Over Time

40 Apple s iphone Algorithmic Analysis of Complex Notion The onboard accelerometer, gyro, and GPS chip help the Apple Watch know how you're moving in the water. Apple built its own algorithms to discern the stroke of swimmers of all skill levels, as well the calories they re burning. The process involved gathering data from more than 700 swimmers and over 1,500 swim sessions.

41 The Human Nervous System Why IoT will be as Revolutionary as FIRE Intelligent Internet of Things

42 Thank You!