HeathCare Sriram Chellappan Dept of Computer Science and Engineering University of South Florida sriramcusfedu httpwwwcseusfedushri Statistics on Mental Health USA 1 in 4 adults experience mental health problems in a given year 75 million people ID: 1044558
Download Presentation The PPT/PDF document "Recent Developments in Big-data Enabled ..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
1. Recent Developments in Big-data Enabled Personalized HeathCareSriram ChellappanDept. of Computer Science and EngineeringUniversity of South Floridasriramc@usf.eduhttp://www.cse.usf.edu/~shri/
2. Statistics on Mental HealthUSA1 in 4 adults experience mental health problems in a given year (75 million people)1 in 17 adults live with a serious condition like schizophrenia, major depressive disorders, bi-polar disorders and moreUpto 30% of college students experience mood problems at some pointAsiaStatistics are much higher for adults for Stress and AnxietyUpto 11% of South Korean youth are reported to be addicted to the InternetConsequencesSuicidal Tendencies, Poor productivity, Economic losses, High medical costs, Unhappy families and more
3. Statistics on Internet Usage
4. Statistics on Mobile Web
5. Presentation OverviewBig-data clues on Mental Health and AlgorithmsMethodology for extracting real and big-Internet dataContinuous, Unobtrusive and Privacy PreservingNovel features to study mood disordersLearning algorithms to predict mood disorders from passive Internet usage monitoringApp development and CommercializationReal challenges faced by the mental health communityOverview of app Future workWearables and HealthCare
6. Part 1 – Big-data Clues and Algorithms
7. State of the ArtInternet data collected via surveys – self reported usageHow many time did you check email last weekHow many times did you chat last weekHow many hours did you use facebook last month…But as we knowInternet data is high dimensional, large volume and multi modalHence quantity, dimensionality and granularity of data collected via surveys is very limitedConclusions are naturally generalized and may have limited practical applicability
8. Advancing State of the Art ??? Continuous, Unobtrusive and Privacy Preserving Collection
9. NetFlow – Developed by Cisco
10. NetFlow…Detect network anomaliesScanning for suspicious port activityEfficient troubleshootingResources deployment…
11. First of All - Internet Data Collection
12. Our Internet dataFor each subject, we now haveBut, Internet usage of a student over weeks runs into millions of records – GB sizeStart TimeEnd TimeDest-ipProtocolSrc PortDest PortOctetsPacketsDuration10/1/12 13:1210/1/12 15:12164.33.22.366504580200155010/2/12 1:1210/1/12 1:4487.34.33.266505380323239810/2/12 16:1210/2/12 17:11123.44.3.221765082443123445129………………………
13. Experiment 1 – Depressive SymptomsWe recruited 216 UG students to participate in our study in Feb 2011Students completed the Center for Epidemiologic Studies Depression (CES-D) scale30% met the minimum criteria for depressive symptoms
14. Experiment 2 – Internet AddictionIt is a relatively new phenomenon and characterized as behavioralRecent studies indicate that Upto 3.5% of German teens are reported to be addictive Internet usersUpto 8.5% of teens in Greece report addiction to Internet with males and gaming dominating the studyUpto 11% of South Korean youth are reported to be addicted to the InternetUnfortunately, we don’t have clear norms yetMany clinics are now operating to treat Internet Addiction across the worldOur survey with 69 students in Fall 2012Problematic Internet Usage Questionnaire (Monash U.)Introversion, Craving, Tolerance, Reduced Activities
15. Experiment 3 – Stress and Anxiety70 students across departments participated in the study in Fall 2012Students were given Depression, Anxiety and Stress (DASS) Questionnaire questionnaire to assess Stress and Anxiety14 students met the criteria for Stress and 7 students met the criteria for Anxiety
16. Features ExtractionWe processed NetFlow records to derive three broad categoriesAggregate Features – Summarize raw aggregates of Internet usage
17. Features Extraction – Cont’dApplication Features – Summarize aggregates of Internet usage categorized by application type
18. Features Extraction – Cont’dEntropy FeaturesDifficulty concentrating or making clear decisions is a symptom of mood disorders, especially depressionWe wanted to capture this behavior from Internet usageShannon Entropy of 1) Source IP address, 2) Destination IP address, 3) Source port, 4) Destination port, 5) Protocol, 6) Octets, 7) Packets and 8) Duration
19. Summary of Critical Statistical Results Packets per Flow Chat Octets Peer to Peer Packets Peer to Peer Octets Mail Duration FTP duration Flow Duration Entropy Chat Duration Chat Octets Streaming Duration Total Duration Mail Duration Total Duration Total Number of Flows Gaming, Chatting, Email Octets Packets per Flow Flow Duration Entropy Depressive SymptomsStress and AnxietyTolerance, Craving, Withdrawal, Reduced Activities
20. Algorithms for Classification of Mood DisordersWe are currently implementing machine learning algorithms to classify mood disorders from run time Internet usage monitoringResults with Support Vector Machines are pioneering and highly encouragingApplications are enormousPrecisionRecallAccuracyDepressive Symptoms88%86%87%Stress91%87%89%Anxiety87%87%87%
21. Wrapping up Part 1Internet usage can reveal significant information about a person’s behaviorBeing ‘big’ and ‘high-dimensional’, challenges are there in collecting and processing data setsOne can extend our study to identify more general and more fine-grained features as well – there is still a high degree of privacyPotential Utility – Orange Societies from Internet UsageProactive and Personalized Mental Health careCare for young and elderly whose problems are not diagnosedCare in rural areas where services are sparseEven Cyber Security…
22. Media Coverage
23.
24. Part 2 – App Development for Orange Societies91 million in USAdults live in areas affected by shortages of mental health workers51 percentOf all US counties have no psychiatrists, psychologists, or social workersNo AccessSome Access
25. State of the Art+Mood
26. Opportunities
27. Mental Health App CategoriesScreeningReferenceTherapy DeliveryDrug Reminders…
28. Fully encrypted and meeting all HIPAA Standards
29. Our App – Mood-Trek on Android
30. Mood-Trek…
31. Mood-Trek…
32. Mood-Trek
33. The Physician View
34. Performance MetricsHeathCare MetricsFaster Turnaround time in CureReduction in PolypharmacyMinimize visits to PhysiciansReduced visits to ERTech MetricsStandard testing, debugging, and sentiment analysis
35. Where we are now
36. Wearable, BigData and HealthCareActivity Recognition is an area of active research in the Tech and HealthCare domainsIndustry and Academia are aggressiveAccelerometer, Gyroscope, Ambient Light, Altimeter, Heart Rate and moreRecent hardware advances also enable Cortisol monitoring, Blood ColorThe field is exploding…
37. Our Work on Activity ClassificationProblem Statement – Detecting complex human activities using multi-positional sensors in body, and ambient Bluetooth beaconsSensors in waist, back, leg, and wrist positions are usedMini Bluetooth beacons provide location contextFeature Extraction from Accelerometer, Gyroscope (from body), Temperature, Pressure and Humidity (from ambience) SensorsMachine learning algorithms
38. Rationale
39. Results
40. HealthCare RelevanceMedicare Access and CHIP Reauthorization Act of 2015 (MACRA) - what this means in the context of Coronary Heart DiseaseActivities Recognition is Vital – Called Ecological Momentary AssessmentsThen comes JITAI – Just in Time Adaptive InterventionsMany companies are entering this space in different forms – Moving Analytics. Wellness mate
41. Wrapping it upThe field is congested, but there are clear opportunities for innovation
42. AcknowledgmentsNational Science Foundation CAREER, SaTC, REU and CRI programsArmy Research OfficeNational Security AgencyDept. of EducationUniversity of South Florida and University of Missouri System
43. Thank you