Ubiquitous Computing and Agent-Based Medical Applications Thomas M

Ubiquitous Computing and Agent-Based Medical Applications Thomas M. Gatton Professor, National University San Diego, California USA

PRESENTATION OVERVIEW • • • Overview of Ubiquitous Computing Current Research Agent-Based Medical Applications and Ubiquitous Computing Automated Knowledge Acquisition Human Disease Diagnosis and Treatment Conclusions and Recommendations

Evolution from Mainframe to Wireless • Herman Hollerith and IBM World Domination – Relays to Vacuum Tubes • Competing Mainframes – Silicon Transistor Revolution • Minicomputers • Microcomputers • • • – Gates, Jobs, et al Cell Phones Internet Wireless PDA’s and WIFI Other Hardware and Software Technologies – – – Sensors, GPS and RFID’s AI Feedback and Agents • Neural Nets, Data Mining and Web Bots

Definitions Ø Mark Weiser Ø 1952 -1999 Ø Ubiquitous Computing ØUbiquity is the ability to Øbe present everywhere or Øat several places at once ØPervasive Computing ØEveryware ØCalm Computing ØAmbient Informatics ØFrom Latin Ubique -> Ubiquidous

Weiser’s vision: Ubiquitous Computing • Computers everywhere, disappearing/integrated in environment/objects around us • Computer no longer isolates us from tasks/environment, no longer focus of attention • Social Impact – Similar to writing: found everywhere from clothes labels to billboards – Similar to electricity which surges invisibly through the walls of every home, office, car • Ubiquitous computing constitutes a reversal of some other trends • Ubiquitous computing does not mean: – Computers that can be carried everywhere – Multi-media computers (using more sensors/output modalities) – Virtual reality (create a world inside the computer, rather than enhance the real world with computer data) – Computer as personal assistant, “agent” • Ubiquitous Computing – – Hundreds of computers in every room Wirelessly networked With their own display Computation happens in the background

Ubi. Comp Techno Lingo • • • Radio tags, smartcards and embedded computers Models of context and context awareness Location models and location awareness Wireless networks and mobile computing Wearable computing Perception and computer vision Human-computer interaction Smart environments and smart spaces Physical-virtual integration Security and privacy Wake University Definitions – Ubiquitous Computing—all members of the academic community have appropriate and timely access to the Internet, usually via a computer they own. Access may be either by desktop or laptop or handheld. – Portable Computing—same as #1 except computer must be a laptop – Mobile Computing—same as #2 except laptop must be wireless. – Very Mobile Computing—same as #3 (i. e. wireless) except that the computer is a Palm Pilot or Blackberry or Equivalent

Current Research ØUnited States ØGroup for User Interface Research, University of California, Berkeley ØIntel Research Seattle ØMIT Media Lab ØMobile & Pervasive Co ØPARC ØStanford University Interactive Workspaces ØSURG, Indiana University ØUbiquitous Computing Research Group, Georgia Institute of Technology ØUCSD, explorations in community-oriented ubiquitous computing ØDUB Group, University of Washington ØKorea ØACAMUS Group, Autonomous Context-Aware Middleware for Ubiquitous computing Systems ØUbiquitous Computing Lab, Kyung Hee University ØNumerous Others

UCSD Active. Campus Explorer uses a person's context, like location, to help engage them in campus life. UCSD Active. Campus

Medical Applications ØHealth. Pal ØRFID - Radio Frequency IDentification ØFood Intake Monitoring ØCusto. Med

Triage System using RFID

Mobile, Wireless Telemedicine Service System Architecture Wireless Phones PDA’s Portable Computers

Medical Healthcare Monitoring with Wearable and Implantable Sensors

Sensor Devices

RFID Radio Frequency IDentification

The Future? ? External Systems BORG: The ultimate ubiquitous system? Internal Systems: Nano. Bots? ?

HUMANS + MACHINE = ? ? Computer Simulation

Looks Like Professor Jens Pohl

Agent-Based Medical Applications Medical Diagnostic Systems Background Patient specific diagnosis and treatment recommendations accuracy problem Knowledge Acquisition Bottleneck Problem

Medical Diagnostic Systems Background • Artificial Intelligence in Medicine (AIM) • • workshops Disseminated the techniques of artificial intelligence – (Mc. Carthy/AI/Agents) Demonstrated successful applications in early systems – MYCIN: production rules and confidence factors – CASNET: semantic network for knowledge representations – PIP: frame-based structure in its knowledge base – INTERNIST: a tree like taxonomy that described symptoms of each disease

Problems of Development and Application ØPatient specific diagnosis and treatment recommendations accuracy problem Ø Physician Referencing, only ØKnowledge Acquisition Bottleneck Problem Ø Traditional Methods Very Slow Ø “Shell” Systems ØSOLUTION – Rapid Development of Customer Specific Knowledge Bases

Automated Knowledge Acquisition • Background • Data Mining and Machine intelligence approaches – Diagnostic Accuracy and Physician Accountability Issues – Explanation facilities limitations • Rule-Based Systems – Knowledge Structure – Decisions trees • Direct knowledge Acquisition from Experts • Systems Modeling and Diagnostic Knowledge Acquisition – Graphical Abstract Domain Model – Extension to Operations and Maintenance

Graphical Abstract Domain Model Engine Block Starter Generator Battery Water Pump

GADM and Symptomatic Links Decision Tree Generation ELECTRICAL SYSTEM STARTER CAR WON’T START BATTERY LOW LIGHTS ARE DIM STARTER CLICKS HEADLIGHTS LIGHT OFF

Human Disease Diagnosis and Treatment • Domain of Diabetes – Occurring at accelerating pace – the failure of the body to regulate blood glucose level • Diabetes Symptoms – The symptoms of diabetes are caused by hyperglycemia, or high blood sugar, and include frequent hunger, urination and thirst, blurry vision, fatigue, dry mouth, slow wound healing, infection susceptibility, loss of weight, dizziness, confusion, weakness, and tremors and, in males, erectile dysfunction. In its more serious stages, diabetic ketoacidosis can occur and is evident by nausea, vomiting and abdominal pain. Patients with diabetic ketoacidosis can quickly go into shock, coma, and even death without prompt medical attention • Diabetes is classified as either type 1, type 2, gestational or other

Ubiquitous Diabetic System Scenario • Implantable devices, such as Radio Frequency • • Identification Tags (RFID) will monitor the internal conditions of the human body, such as glucose levels, heart rate and blood pressure, and wirelessly communicate to agent based systems for analysis, diagnosis and real-time medical treatment Other monitoring devices can be used to gather information about caloric intake, weight and activity levels, and use wireless technologies, such as PDA’s, to transfer this data to appropriate locations for evaluation and action. While monitoring devices may trigger alarms, action is usually by physicians, nurses and other appropriate medical personnel who may consult medical knowledge based systems for confirmation

Classification of Diagnosis, Treatment and Monitoring • Human Body as a System • Diabetes treatment consists primarily of weight control, exercise, • • diet and medication. Complicated by specific patient characteristics, such as dyslipidemia in children and adolescents, postprandial blood glucose, foot wounds, coronary heart disease, insulin resistance, nephropathy, lipid disorders, hypertension, cardiovascular risk factors Can lead to disability, death and other diseases, such as acute metabolic complications, vision disorders, neuropathy, kidney dysfunction, peripheral vascular disorders, lower extremity foot ulcers and amputations, heart and digestive diseases, infections and oral and psychosocial complications

Treatment and Monitoring as Operations, Maintenance and Diagnosis • The patient has the primary responsibility for following the • • • treatment plan. It will include the monitoring of blood glucose levels and awareness of the occurrence of symptoms indicating hyperglycemia. The collection of data documenting the patient’s caloric intake, exercise, blood pressure and glucose levels assists evaluation of the treatment plan to determine its effectiveness. When the patient’s symptom(s) fall outside of a given range, there is a diagnosis and recommendation for that condition. The specific recommendation may include changes in medication, diet, office visits or exercise as well as emergency medical action. Each recommendation is dependent on the degree of departure from normal or acceptable symptomatic limits determined by the physician for a particular patient’s condition.

Direct Expert Knowledge Acquisition Algorithm Diabetic Diagnosis and Treatment Plan Patient Ubi Syst Existing Conditions, Symptoms and Tests Results Physician Diabetic Treatment Knowledge

Ubiquitous Healthcare System Organization Treatment Plan Patient Symptoms Patient Actions Ubiquitous Agent Based System Patient Condition

AKA Algorithm Start Input Treatment Plan Symptoms Are Known Input Symptoms and Conditions States Generate Symptom Condition State Generate Rule More States? Finish

Conclusions and Recommendations Questions? ? Thank-You