Presentation is loading. Please wait.

Presentation is loading. Please wait.

Programme Committee Chair Silvia Miksch, Vienna, Austria Organising Committee Chair Jim Hunter, Aberdeen, UK Artificial Intelligence in Medicine.

Similar presentations


Presentation on theme: "Programme Committee Chair Silvia Miksch, Vienna, Austria Organising Committee Chair Jim Hunter, Aberdeen, UK Artificial Intelligence in Medicine."— Presentation transcript:

1 Programme Committee Chair Silvia Miksch, Vienna, Austria Organising Committee Chair Jim Hunter, Aberdeen, UK Artificial Intelligence in Medicine

2 Where... AIME 05 was hosted by the Department of Computing Science, University of Aberdeen

3 Participants Australia2 Austria9 Canada5 Cyprus1 Czech Rep.1 Denmark4 France 19 Germany6 Greece1 Hungary1 Italy 15 Israel1 Lithuania2 Netherlands 12 Nigeria2 Poland1 Russia1 Slovenia3 South Korea3 Spain4 Sweden3 Switzerland2 Thailand2 UK 13 USA8 42% students 120 registrations for main conference 16 for pre-conference events

4 New Event Doctoral Consortium Chair: Elpida Keravnou (Nicosia, Cyprus) 8 PhD Students Participating Faculty Ameen Abu-Hanna (Amsterdam, Netherlands) Riccardo Bellazzi (Pavia, Italy) Carlo Combi (Verona, Italy) Michel Dojat (Grenoble, France) Peter Lucas (Nijmegen, Netherlands) Silvana Quaglini (Pavia, Italy) Yuval Shahar (Beer Sheva, Israel)

5 Tutorials Evolutionary Computation Approaches to Mining Biomedical Data John Holmes (University of Pennsylvania, USA) Causal Discovery from Biomedical Data Subramani Mani (University of Pittsburg, USA) Applied Data Mining in Clinical Research John Holmes (University of Pennsylvania, USA)

6 Workshops IDAMAP-2005: Intelligent Data Analysis in Medicine and Pharmacology Niels Peek (University of Amsterdam, Netherlands) John Holmes (University of Pennsylvania, USA) Biomedical Ontology Engineering Jeremy Rogers, Alan Rector, Robert Stevens (University of Manchester, UK)

7 Submissions and Reviewing 2005 – Reviewing Process PC Members: + 13 persons incl. 5 physcians additional reviewers approx. 3 reviews / paper 128 % more submissions

8 Acceptance Rates 2005: 35 long 33 short AIMDM 99 = AIME + ESMDM

9 Accepted Papers by Country

10 Invited Talks Frank van Harmelen (Vrije Universiteit, Amsterdam, Netherlands) Ontology Mapping: A Way out of the Medical Tower of Babel?

11 Different approaches to ontology matching Linguistics & structure Shared vocabulary Instance-based matching Shared background knowledge

12 Conclusions Ontology mapping is (still) hard & open Many different approaches will be required: linguistic, structural statistical semantic … Currently no roadmap theory on what's good for which problems

13 Invited Talks Paul Lukowicz (University for Health Sciences, Medical Informatics and Technology, UMIT Hall in. Tirol, Austria) Human Computer Interaction in Context-Aware Wearable Systems

14 Wearable Vision Wearable Vision non disruptive interaction environment oriented output environment oriented context recognition/monitoring physically unobtrusive technology systems seamlessly connected wear. system user real world 100% > > 5 0 % % applications

15 What is Context Recognition ? Embedded Controllers: feedback control loop Artificial Intelligence: imitating human cognition and perception - includes interpretation Context Recognition: mapping signals from a group sensors onto a set of predefined, environment related states

16 Wearable Vision Wearable Vision wear. system user real world 100% > > 5 0 % % wearable computer: system as an enhancer and facilitator in the interaction between the user and the real world a whole new field of applications context recognition is the key issue

17 Temporal Representation and Reasoning Decision Support Systems Clinical Guidelines and Protocols Ontology and Terminology Case-Based Reasoning, Signal Interpretation, Visual Mining Intelligent Image Processing Knowledge Management Machine Learning, Knowledge Discovery and Data Mining Programme - Sessions

18 Temporal Rep. & Reasoning Topics Temporal Data Abstraction Temporal Patterns, Probabilistic Methods Learning Temporal Rules Temporal Data Models Fuzzy Temporal Framework Bayesian-Network Models Application Areas (N)ICUs: Artificial Ventilation, Blood Glucose Hemodialysis Sessions, Gene Expression Data

19 Rules [Antecedent, Concequent] Temporal Relations Two Episodes PRECEDES: Allens Temp. Operators Applications Hemodialysis Sessions Gene Expression Data Learning Rules with Complex Temporal Patterns in Biomedical Domains Lucia Sacchi, Riccardo Bellazzi, Cristiana Larizza, Riccardo Porreca, Paolo Magni Time series represented through complex TAs I = [Increasing] D = [Decreasing] S = [Steady] Definition of a set of significant patterns P = {p 1, …, p N } P 1 = [Increasing Decreasing] P 2 =[Decreasing Increasing] INPUT: raw data (biomedical time series) APRIORI-like rule extraction algorithm OUTPUT: set of temporal rules TA mechanism Time series represented through a set of basic trend TAs a) b) c)

20 Support (Sup) = RTS / TS Confidence (Conf ) = NARTS / NAT Example: Hemodialysis Sessions Monitoring Learning Rules with Complex Temporal Patterns in Biomedical Domains Lucia Sacchi, Riccardo Bellazzi, Cristiana Larizza, Riccardo Porreca, Paolo Magni

21 Point-based Semantic Problems Downward Inheritance Upward Inheritance Countability A Three-Sorted Model: Interval-based Semantic Solution Downward Inheritance Upward Inheritance Countability Extending Temporal Databases to Deal withTelic/Atelic Medical Data Paolo Terenziani, Richard Snodgrass, Alessio Bottrighi, Mauro Torchio, Gianpaolo Molino Data Models Grounded...

22 Ceilidh: Scottish Country Dancing

23 Clinical Guidelines & Protocols Topics Design Patterns for Modelling CGPs Information Extraction for Modelling CGPs CGP Representation, Execution & Adaptation Usability of CGPs User Interfaces for CGPs Recommendations Decision Theory for CGPs Selection CGP Retrieval – Concept Hierarchies Quality Indicator for CGPs Application Areas Asthma, Diabetes, Jaundice, (N)ICUs, Oncology, Otolaryngology (CGP)

24 m Architecture Testing Asbru Guidelines and Protocols for Neonatal Intensive Care Christian Fuchsberger, Jim Hunter, Paul McCue Data Abstraction Execution Engine Test Data Visualisation Guideline Recommendations IF O2 > O2-High THEN Rec_FiO2 = FiO2 - 5 IF O2-High> O2 > O2-Low THEN Rec_FiO2 = FiO2 IF O2-Low > O2 THEN Rec_FiO2 = FiO O2-High O2-Low O2 FiO2 - FiO2 + 8 kPa 6 kPa Example: CGP: O 2 Management Results

25 The Problem Gaining Process Information from Clinical Practice Guidelines Using Information Extraction Katharina Kaiser, Cem Akkaya, Silvia Miksch Knowledge-intensive Cumbersome Time-consuming Automation Structuring Traceability

26 Information Extraction Process Information Gaining Process Information from Clinical Practice Guidelines Using Information Extraction Katharina Kaiser, Cem Akkaya, Silvia Miksch Results Task 1: Detecting relevant sentences Filtering and segmentation module Recall: 76 % Precision: 97 % Task 2: Extracting processes Process extraction, merging & grouping modules Recall: 94 % Precision: 84 %

27 Idea: Linguistic Patterns Ontology-Driven Extraction of Linguistic Patterns for Modelling Clinical Guidelines Radu Serban, Annette ten Teije, Frank van Harmelen, Mar Marcos, Cristina Polo-Conde instance[radiotherapy, produces, skin_reactions] inst_of template[med_action, effect_op, med_effect] covers o_fragment(MedAction produces MedEffect) MedActionMedEffect produces radiotherapy skin_reactions

28 Evaluation Linguistic Patterns Breast cancer GL (chapters 2-4) Ontology-Driven Extraction of Linguistic Patterns for Modelling Clinical Guidelines Radu Serban, Annette ten Teije, Frank van Harmelen, Mar Marcos, Cristina Polo-Conde

29 Audience …

30 Ontology and Terminology Topics Design & Building a (Domain) Ontology Ontology of Time & Situoids Terminology Extraction from Text Terminology Alignment Population Ontology Using NLP & ML Application Areas Allergens, Oncology, Surgical ICUs, ICUs, Tissue Microarrays

31 The Problem Combining different patient registrations Terminologies have to be mapped Ontology mapping is hard problem in general Especially when terminologies contain no structure Situation ACM, OLVG: list of reasons for ICU admission DICE: hierarchical knowledge describing the reasons for ICU admission Using Lexical and Logical Methods for the Alignment of Medical Terminologies Michel Klein, Zharko Aleksovski

32 The Approach aspect taxonomies given anatomical location abnormalitybody system 1. lexical methods 2. classification aneurysmaaorta acutepharyngitis aneurysmacordis axillo-poplitealebypass cerebrovasculair accident coloncarninoom … AIDS acutepancreatitis aneurysmavan aorta arterialhaemorrhage … tuberculeuzemeningitis tumor … structured ontologylist of terms Using Lexical and Logical Methods for the Alignment of Medical Terminologies Michel Klein, Zharko Aleksovski OLVG: Acute respiratory failure DICE: Asthma cardiale OLVG: HIV DICE: AIDS OLVG: Aorta thoracalis dissectie type B DICE: Dissection of artery cause location, abnormality abnormality Example Results

33 … still there!

34 Intelligent Image Processing Topics Electrocardiographic Imaging – Activation Maps Cellular Neural Networks Sketch Understanding Automatic Segmentation Support Vector Machines Application Areas ECGs Analysis, Cephalometric Analysis – X-rays Anatomy, Bone Scintigraphy (Whole-body Bone Scan)

35 Idea: Anatomical Sketching Sketching is ubiquitous in medicine Patient records Communication with patients Consultation Medical education Anatomical Sketch Understanding: Recognizing Explicit and Implicit Structure Peter Haddawy, Matthew Dailey, Ploen Kaewruen, Natapope Sarakhette

36 Anatomical Structure & View … Template 1 Template m Naïve Bayes Classifier Lung External User Sketch Template 1 Template m … Shape Context Matching Recognition Process

37 Data Collection Collected sample sketches from 3 rd – 6 th year medical students 70 sketches of 6 anatomical structures, 2-3 views per structure: 1050 sketches Compared: Student accuracy vs UNAS (UNderstaning Anatomical Sketches) accuracy Anatomical Sketch Understanding: Recognizing Explicit and Implicit Structure Peter Haddawy, Matthew Dailey, Ploen Kaewruen, Natapope Sarakhette

38 Evaluation: Accuracy Anatomical Sketch Understanding: Recognizing Explicit and Implicit Structure Peter Haddawy, Matthew Dailey, Ploen Kaewruen, Natapope Sarakhette Baseline random classification accuracy = 6.7%

39 Conference Dinner

40 Knowledge Management Topics Multi-Agent Patient Representation Clinical Reasoning Learning Process Reengineering Application Areas Primary Care Cognitive Processes during Clinical Reasoning Cardiac Infarction Diagnosis Hospital Logistic Processes

41 Machine Learning,Knowledge Discovery & Data Mining Topics Web Mining Clustering Methods: Similarities & Statistical Techniques Naïve Bayesian, Rule-based, Case-based, Tree-based, and Genetic Algorithm, Inductive Logic Programming Scenarios Learning Subgroup Mining – User-Guided Refinement Application Areas Acute Paediatric Abdominal Pain, Cardiac Monitoring, Corneal Disease, Dental Medicine (Prosthetic Appliance), Dietary Menu Planning, Epidemiologic Surveillance, Gene Expression Data, ICU, Influenza Sequences, MR Spectra, Oncology, Public Health Care

42 User-Guided Approach Subgroup mining method => potentially guilty (faulty!) elements Visualization to ease interpretation User has full control of the refinement process Subgroup Mining for Interactive Knowledge Refinement Martin Atzmueller, Joachim Baumeister, Achim Hemsing, Ernst-Jürgen Richter, Frank Puppe

43 Visualization Subgroup Mining for Interactive Knowledge Refinement Martin Atzmueller, Joachim Baumeister, Achim Hemsing, Ernst-Jürgen Richter, Frank Puppe

44 Conclusions AIME is healthy 128 % more submissions than last AIME-2003 => Decrease of acceptance rate (long: 23.6%; short: 22.3%) General Highlights Doctoral Consortium 3 Tutorials and 2 Workshops Two excellent invited talks

45 Conclusions Content Highlights Medical application areas are very broad Clinical Guidelines and Protocols has matured Ontologies and Terminologies is a hot topic and generated a lot of discussion Dealing with temporal data and information is crucial Comparing the usefulness of different machine learning and mining techniques brings more insights Intelligent Visualization is emerging in AIME

46 AIME 07 Academic Medical Centre University of Amsterdam Ameen Abu-Hanna Local Organiser: Ameen Abu-Hanna will be hosted by:

47 Haste ye back (Come back again soon!) AIME 05

48


Download ppt "Programme Committee Chair Silvia Miksch, Vienna, Austria Organising Committee Chair Jim Hunter, Aberdeen, UK Artificial Intelligence in Medicine."

Similar presentations


Ads by Google