The document discusses using affordable sensing technologies for data-driven healthcare applications in screening, diagnosis, and therapy. It describes challenges in healthcare access and affordability in developing and developed countries. The TCS HeartSense and RehabBox technologies are presented as using sensors like PPG, PCG and Kinect to enable affordable screening for conditions like CAD, hypertension, and remote rehabilitation for stroke patients. Pilot studies show potential for these technologies to provide early screening and monitoring of diseases. The document outlines several publications and patents from ongoing work in this area.

1. Data-driven Healthcare using Affordable Sensing
- Screening, Diagnosis and Therapy
Arpan Pal
Head, Embedded Systems and Robotics
TCS Research
Tata Consultancy Services, India

2. 2TCS Confidential
Developed Countries
Elderly people - 44.7 M (2013), double
by 2060
Invasive and costly diagnosis
One size fits all Diagnostic / Treatment
protocols
Some diseases yet to have a cure
Developing Countries
Capacity - not enough doctors per
patient
Reachability – specialized primary care
not available
Affordability - majority cannot afford to
pay the cost
http://www.aoa.acl.gov/Aging_Statistics/index.aspx
Problems of the New Age and the New World

3. 3TCS Confidential
Next Generation Health
Keeping People
Healthy
Monitoring
Alert generation
Identifying
Disease Onset
Better
Diagnostics
Predictive CareTreating Full Fledged
Disease
Conventional and New
Care Protocols
Reaction To
Disease
Prevention
of Disease
Disease
Treatment
Promote
Wellness
Now Future Improvements Across the Spectrum from Diseased to Healthy

4. 4TCS Confidential
Application Use Cases
Accurate, Affordable, Accessible,
Adaptive diagnostics & therapeutics
Screening /
Diagnosis
Coronary Artery
Disease (CAD)
Hypertension
Therapy
Physical rehab for Stroke
Patients
Future Plan
• Diabetes
• Chronic Obstructive
Pulmonary Disease (COPD)
• Neurorehab
HeartSense RehabBox

5. 5TCS Confidential
CAD – also a global hazard
IHD [ischemic heart disease, also known as CAD] causes more deaths
and disability and incurs greater economic costs than any other
illness in the developed world. . . [and it] is likely to become the
most common cause of death worldwide by 2020 - Antman et al.,
2008
DoctorSpeak
“Only Existing conclusive diagnostic procedures is coronary
angiogram - invasive, potentially harmful, costly.
It is important to diagnose this preventive condition of CAD
with affordable non-invasive mass screening.”
Dr. K M Mandana - Consultant, Fortis Hospital, Kolkata
MCh (Cardiovascular & Thoracic Surgery), Fellow of European
Board of Cardiothoracic Surgeons
CAD – Motivation
Need for Early Screening by detecting/predicting early onset of CAD

6. 6TCS Confidential
Singapore–10, USA -14, UK - 9
Hypertension – Motivation
2011 2012 2013 2014
206 197 189 181
India – Maternal Mortality Rate per 100K Live Berths
data.worldbank.org/indicator/SH.STA.MMRT
Causes of maternal deaths
1. Direct Causes: 81%
Severe Bleeding 25%
Sepsis 15%
Unsafe abortions 13%
Eclampsia 12%
Obstructed Labour 8%
Other direct causes * 8%
2. Indirect Causes ** 19%
* Other direct causes: Ectopic pregnancy, embolism, anesthesia related.
** Indirect Causes: Malaria, Anemia, Heart Diseases.
Rehana Kausar, “Maternal Mortality in India - Magnitude, Causes and Concerns”, Indian Journal for the Practising Doctor, Vol. 2, No. 2 (2005-05 - 2005-06)
http://www.indmedica.com/journals.php/mailt?journalid=3&issueid=58&articleid=722&action=article
Eclampsia - a condition in which one or more
convulsions occur in a pregnant woman
suffering from high blood pressure, often
followed by coma and posing a threat to the
health of mother and baby.
Need for Early Screening of Hypertensive Mothers – control of Hypertension by medication significantly
reduces complications
Also need for Hypertensive Screening in general
Hypertension – also a global hazard
USA Figures
• Hypertension - leading cause of heart disease and stroke
• 1/3rd have hypertension, another 1/3rd prehypertension
• Only 54% of hypertensive people have it under control
• Total cost to Healthcare - US$48.6 billion each year.
http://www.cdc.gov/dhdsp/data_statistics/fact_sheets/fs_bloodpressure.htm

7. 7TCS Confidential
HeartSense Technology
Systolic Peak
Diastolic Trough
Acquisition and Pre-
processing
Feature Engineering
and Extraction,
Cardiovascular
Models
Classification
(includes metadata
like age, gender,
height, weight, BMI)
Photoplethysmogram (PPG) Phonocardiogram (PCG)

8. 8TCS Confidential
TCUP - TCS IoT Platform
Ready and DeployedOngoing Field Trials
Photodetector or Camera Microphone
PPG
Cardiovascular
Models
PCG
Expert Doctor
Real-time View
CAD Screening
Event Alerting
Maternal Hypertension
Screening
HeartSense Architecture
Advanced Analytics
Cardiovascular
Models

9. 9TCS Confidential
CAD – Pilot Study Results
CAD
- Own Collected dataset (20 non-CAD + 15
CAD)
- 15 healthy adults with no known
history of cardiac problem at TCS
Kolkata
- 5 patients but non-CAD subjects from
Fortis)
- 15 angio-proven CAD from Fortis
- MIMIC-II matched dataset (ICU)
- 56 CAD and 56 non-CAD
Sensitivity = correctly identify CAD
Specificity = correctly identify non-CAD
Own Collected Data MIMIC – II Data
Specificity Sensitivity Specificity Sensitivity
PCG 80% 60% - -
PPG 40% 90% 82% 88%
Fusion 70% 80% - -
 1000 patient trial in two hospitals in India – Fortis Hospital in Kolkata and JJ Hospital in Mumbai
 IRB clearance in place, MoU in progress
Future Plan

10. 10TCS Confidential
Blood Pressure – Pilot Study Results
BP
98 subjects at 2 hospitals (Narayana Nethralaya, Bangalore
and Primary healthcare centre, Jamla, Gujrat)
- SBP (80-210 mmHg) and DBP (60-130 mmHg)
Error
(mmHg)
SBP DBP
OMRON1 iBP2 TCS Model OMRON1 iBP2 TCS Model
< 5 61% 24% 29% 52% 26% 41%
< 10 85% 44% 43% 85% 48% 66%
< 15 94% 59% 66% 96% 70% 84%
• Working with Municipality and Government in Nashik, a semi-urban eco-system
• Focusing more on classification for Hypertensive screening and not on actual BP measurement – MAP based
• Field Trial with Pregnant mothers planned – need for Longitudinal Study
Future Plan
1. El Assaad, Mohamed A., et al. "Validation of the Omron HEM-907 device for blood pressure measurement." American Journal of Hypertension 15.S3 (2002): 87A-87A.
2. Plante, Timothy B., et al. "Validation of the Instant Blood Pressure Smartphone App." JAMA internal medicine (2016) – A study by Johns Hopkins University School of Medicine

11. 11TCS Confidential
HeartSense - Achievements so far
Contribution
• Low cost Mobile
Attachment / wearable
• Digital stethoscope
• Smartwatch
• Robust Signal Processing
• Outlier Rejection
• Motion Artefact Removal
• Robust Pre-processing
• Cloud based screening and
alerting
• Supervised Learning for
CAD, Hypertension
Classification
• Multi-sensor Fusion
IP
• Papers
•15+ in major conferences
(ICASSP, EMBC, MobiHoc,
Sensys, ICC, BSN ..)
• Patents
•14 patents filed
• Collaborations
•Indian Statistical Institute
Kolkata
•Fortis Hospital, Kolkata
•IIT KGP and IISc planned
Awards
• Awards
•Wearable Tech Award, 2016
•Aegis Graham Bell Award,
2015
•CSI Young Innovator Award,
2015
•Best Demo Award at Sensys
2014
Also working on Diabetes, COPD and Stress

12. 12TCS Confidential
Tele-rehabilitation for Stroke Patients - Motivation
Very expensive devices and high
maintenance cost
Existing Quantitative Motion Analysis
systems (VICON ) costs approx. $150K – very
few hospitals in have it even in metro cities
– no access in rural locations
Transportation and Queue at
Hospital
No adaptability in exercises and
Poor Compliance Checking

13. 13TCS Confidential
TCS RehabBox which will be used for balance, gait and joint range of motion analysis using Kinect sensor
Store Raw
Data Patient’s Exercise
Parameter
Patient History
Extract
Parameters
Doctor’s Portal
Feedback
Video Conf.
RehabBox - Architecture
• Single Limb Standing (SLS) for Balance – Duration, Vibration,
Fall Risk Score
• Gait Analysis – Dynamic Balance, Foot Landing, Stride
Length
• Range of Motion (ROM) Analysis for Upper Limb – Joint
movement compliance

14. 14TCS Confidential
RehabBox –Pilot Study Results
Single Limb Stance - Balance
• 11 subjects (5 chronic stroke-survivors, 3 females, mean age 64.4 ± 8.2 yrs) - able to stand/walk unaided.
History of fall was present in 80% stroke and 16.6% control subjects.
• SLS duration (SLSD) was significantly low in Stroke vs Control (SLSD =9.5±14.5 sec. vs 55.7±13.6 sec.).
• Both SLSD and Vibration Index (VI) are significantly different in patients with fall vs no-fall history
(SLSD=6.7±9.8 second vs 59.9±13.8; VI=0.68±0.34 vs 0.25±0.16
• 8 healthy subjects (age: 21-36 years, weight: 47kg - 78kg & height: 4’6’’ - 5’9’’, athletic and non-athletic)
• Minimum Absolute Error (w.r.t. GAITRite) for stride length 3.84 cm
ROM Analysis – Improvement in Kinect Joint Motion Estimate
Future Plan – Larger Patient Trials at AMRI Kolkata and Kokilaben Hospital, Mumbai
Gait Analysis – Stride Length

15. 15TCS Confidential
RehaBox - Achievements so far
Contribution
• Image Processing
• Improvement in Kinect Accuracy
• Joint movement model based
RGBD processing to augment
Kinect Skeleton Model
• Analytics
• Cloud based therapy planning
and compliance
• Accurate Stride length via Gait
Modeling and Analysis
• Creation of a Strength and
Stability Score fusing SLS
Duration and Vibration Index
• Correlate with Fall Propensity
IP
• Papers
•5+ in major conferences (ICASSP, EMBC,
BioMed, ESPRM, INREM..)
• Patents
•5 patents filed
• Collaborations
•AMRI Hospital, Kolkata
•IIT Gandhinagar
Also working on Neurorehab using EEG and GSR

16. 16TCS Confidential
1. Pal, Arpan et. al., "A robust heart rate detection using smart-phone video." In Proceedings of the 3rd ACM MobiHoc workshop on Pervasive wireless healthcare, pp. 43-48. ACM,
2013.
2. Ghose, Avik et. al., "UbiHeld: ubiquitous healthcare monitoring system for elderly and chronic patients." In Proceedings of the 2013 ACM conference on Pervasive and ubiquitous
computing adjunct publication, pp. 1255-1264. ACM, 2013.
3. Banerjee Rohan et. al., "PhotoECG: Photoplethysmography to Estimate ECG Parameters" In IEEE International Conference on Accoustics, Speech and Signal Processing, 2014
4. Dutta Choudhury Anirban et al. "Estimating Blood Pressure using Windkessel Model on Photoplethysmogram" in EMBC 2014.
5. Visvanathan Aishwarya et al. "Smart Phone Based Blood Pressure Indicator" in MobileHealth workshop of Mobihoc 2014
6. Visvanathan Aishwarya et al. "Effects of Fingertip Orientation and Flash Location in Smartphone Photoplethysmography“, ICACCI 2014
7. Banerjee, Rohan et al. "Demo Abstract: HeartSense: Smart Phones to Estimate Blood Pressure from Photoplethysmography" in 11th ACM Conference on Embedded Networked
Sensor Systems (SenSys 2014)
8. Dutta Choudhury, Anirban et al. "HeartSense: Estimating Heart rate from Smartphone Photoplethysmogram using Adaptive Filter and Interpolation" in 1st International Conference
on IoT Technologies for HealthCare (HealthyIoT, IoT-360)
9. Nasim Ahmed et al. “Feasibility Analysis for Estimation of Blood Pressure and Heart Rate using A Smart Eye Wear”, in WearSys Workshop, Mobisys 2015
10. Aditi Misra et al. "Novel Peak detection to estimate HRV using Smartphone Audio“ in BSN 2015
11. "Noise Cleaning and Gaussian Modeling of Smart Phone Photoplethysmogram to improve Blood Pressure Estimation" in ICASSP 2015
12. Keshaw Dewangan, Arijit Sinharay, Parijat Deshpande, “Design And Simulation Of A Low-cost Digital Stethoscope” in COMSOL 2015 as poster for 30 Oct 2015
13. Shreyasi Datta et al, “Blood Pressure Estimation from Photoplethysmogram using Latent Parameters” in IEEE ICC 2016
14. "Rohan Banerjee et al, “Non Invasive Detection of Coronary ArteryDisease using PCG and PPG Signal” in HealthWear 2016 : 1st EAI International Conference on Wearables in
Healthcare"
15. Rohan Banerjee et al, "Time-Frequency Analysis of Phonocardiogram for Classifying Heart Disease", Computing in Cardiology 2016
16. Kingshuk Chakravarty et. al. ,“A Multimodal Therapy Design Toolbox for Gait-Rehabilitation". INEREM 2015
17. Chakravarty, Kingshuk et. al. ,"Quantification of balance in single limb stance using kinect." In 2016 IEEE International Conference on Acoustics, Speech and Signal Processing
(ICASSP), pp. 854-858. IEEE, 2016
18. Kingshuk Chakravarty et. al. ,"AN AFFORDABLE GAIT AND POSTURAL BALANCE ANALYSIS SYSTEM USING KINECT FOR REHABILITATION." In BioMed 2016
19. Sanjana Sinha et. al. ,"Accurate Upper Body Rehabilitation System Using Kinect." Accepted in EMBC 2016 (To Appear in 16-20 Aug 2016)
20. Kingshuk Chakravarty et. al. ,"PREDICTING FALL RISK IN STROKE SURVIVORS: A NOVEL MEASUREMENT USING KINECT BASED SYSTEM", Accepted in World Stroke Congress (To
Appear in 26-29 Oct. 2016)
Relevant Publications

17. 17TCS Confidential
Aniruddha Sinha
Parijat Deshpande
Anirban Duttachaudhuri
Rohan Banerjee
Shreyoshi Dutta
Kingshuk Chakravarty
Brojeswar Bhowmick
Sanjana Sinha
Avik Ghose
Nasim Ahmed
Aditi Misra
Aiswharya Visvanathan
Keshaw Dewangan
Arijit Sinharay
Team

18. Thank You