Machine Learning for Health Care Applications

an ICML 2008 workshop
July 9, 2008, Helsinki, Finland

Health-care applications have been and continue to be the source of inspiration for many areas of artificial intelligence research. Many advances in various sub-specialties of AI have been inspired by challenges posed by medical problems.  A new challenge for AI in general, but machine learning in particular, arises from the wealth and variety of data  generated in modern medical and health-care settings.  Extensive electronic health and medical records---with thousands of fields recording patient conditions, diagnostic tests, treatments, outcomes, and so on---provide an unprecedented source of information that can provide clues leading to potential improvements in disease detection, chronic disease management, design of clinical trials, and other aspects of health-care.  The purpose of this workshop is to bring together machine learning researchers interested in problems and applications in health-care, with the goal of exchanging ideas and perspectives, identifying important and challenging applications, and raising awareness of potential health-care applications in the machine learning community.

The workshop program will consist of presentations by invited speakers, and oral and poster presentations by authors of extended abstracts submitted to the workshop. Confirmed invited speakers for the workshop are: Riccardo Bellazzi, University of Pavia, Italy, and Jeff Schneider, Carnegie Mellon University, USA.

The workshop and invited speakers are supported by AICML.



Contact e-mail:

Workshop program

Room: S5, 3rd floor

9:00-9:50     Invited Talk: Methods and Tools for Mining Multivariate Temporal Data in Clinical and Research Applications
                      Riccardo Bellazzi
9:50-10:10   Machine Learning Techniques in Intensive Care Monitoring
                      Wiebke Sieben, Karen Schettlinger, Silvia Kuhls, Michael Imhoff, Ursula Gather
10:10-10:30 Probabilistic Modeling of Sensor Artifacts in Critical Care
                      Norm Aleks, Stuart Russell, Michael G. Madden, Diane Morabito,
                      Kristan Staudenmayer, Mitchell Cohen, Geoffrey Manley
10:30-11:00 Coffee break
11:00-11:20 Machine Learning to Automate the Assignment of Diagnosis Codes to Free-text Radiology Reports: a Method Description
                      Hanna Suominen, Filip Ginter, Sampo Pyysalo, Antti Airola, Tapio Pahikkal,
                      Sanna Salanter, Tapio Salakoski
11:20-11:40 Conditional Anomaly Detection Methods for Patient-Management Alert Systems
                      Michal Valko, Gregory Cooper, Melissa Saul, Amy Seybert, Shyam Visweswaran,
                      Milos Hauskrecht
11:40-12:00 Bayesian Modelling of Multi-View Mammography
                      Nivea Ferreira, Marina Velikova, Peter Lucas
12:00-12:20 Facilitating Clinico-Genomic Knowledge Discovery by Automatic Selection of KDD Processes
                      Natalja Punko, Stefan RŠuping
12:20-14:30 Lunch
14:30-15:20 Invited Talk: Machine Learning for in vivo Central Nervous System (CNS) Drug Discovery
                      Jeff Schneider
15:20-15:40 Identifying Active Compounds from Chinese Medicinal Plants via Causal Variable Selection
                      Xuewei Wang
15:40-16:00 Optimizing Treatment Strategies for Epilepsy Using Reinforcement Learning
                      Joelle Pineau, Arthur Guez,  Robert D. Vincent, Massimo Avoli
16:00-16:30 Coffee break
16:30-18:00 Poster session
                     Detection of Keratoconus by Semi-Supervised Learning
                      Deepthi Cheboli, Balaraman Ravindran
                     Machine Learning for Personalized Medicine: Will This Drug Give Me a Heart Attack?
                    Jesse Davis, Eric Lantz, David Page, Jan Struyf, Peggy Peissig, Humberto Vidaillet, Michael Caldwell
                     Pattern Discovery in Intensive Care Data through Sequence Alignment of Qualitative Trends Data:
                                  Proof of Concept on a Diuresis Data Set
                      Martijn Devisscher, Bernard De Baets, Ingmar Nopens, Johan Decruyenaere, Dominique Benoit
                     Improving Medical Predictive Models via Likelihood Gamble Pricing
                      Glenn Fung, Harald Steck, Shipeng Yu, Phan Giang
                     Explaining Artificial Neural Network Ensembles: A Case Study with Electrocardiograms from Chest Pain Patients
                     Michael Green, Ulf Ekelund, Lars Edenbrandt, Jonas Bjork, Jakob Lundager Forberg, Mattias Ohlsson
                     Learning Outbreak Regions in Bayesian Spatial Scan Statistics
                      Maxim Makatchev, Daniel B. Neill
                     Detecting Heartbeats in the Ballistocardiogram with Clustering
                      Joonas Paalasmaa, Mika Ranta
                     Classification of Normal and Hypoxic Fetuses using System Identification from Intra-Partum Cardiotocography
                      Philip A. Warrick, Emily F. Hamilton, Robert E. Kearney, Doina Precup
                     Visualisation of High-Dimensional Data for Very Large Data Sets
                     David Wong, Iain Strachan, Lionel Tarassenko

Invited talks

Riccardo Bellazzi, PhD, University of Pavia, Italy.
Methods and tools for mining multivariate temporal data in clinical and research applications.

Abstract. In all human activities, automatic data collection pushes towards the development of tools that are able to handle and analyze data in a computer-supported fashion. In the majority of the application areas, this task cannot be accomplished without using the available knowledge on the domain or on the data analysis process. This need becomes essential in biomedical applications, since medical decision making needs to be supported by arguments based on medical and pharmacological knowledge. It is therefore important to study the computational methods for data analysis aimed to narrow the gap between data gathering and data comprehension, as well as their applications in medicine, health care, biology and pharmacology. Methods for analyzing data by integrating the available knowledge on the domain (Intelligent Data Analysis) and for extracting knowledge from large data-bases (Data Mining) have been investigated over the last few years. In this talk I will deal with the methods for dealing with multivariate temporal data. I will describe in detail two approaches that have been successfully applied in different application problems: the extraction of temporal association rules and the automated construction of dynamic probabilistic models called Dynamic Bayesian Networks. I will show their application in the analysis of hemodialysis monitoring time series, health care administrative data and gene expression data.

Jeff Schneider, PhD, Carnegie Mellon University, USA.
Machine Learning for in vivo CNS Drug Discovery.

Abstract. Researchers in machine learning have made great strides in modeling and optimization of commercial/industrial processes. A more recent trend is to observe that the scientific method is a process that can be modeled and optimized with similar techniques. In this talk we consider a specific example of that: discovery of central nervous system drugs (e.g. antidepressants antipsychotics, anxiolytics, etc.) using in vivo behavioral testing. Algorithms will be discussed in the following areas: the use of kernel density estimators to provide improved posterior probabilities in multi-class applications; the use of semi-supervised learning to handle training data with uncertain class labels; and the use of active learning to control experimentation in the discovery process.

Instructions for full paper submissions

The final papers are due on June 27, 2008. The papers should be at most eight pages long and follow the ICML submission format for final papers. The ICML formatting instructions can be found here. Please email your submissions directly to Csaba Szepesvari, a co-chair of the workshop, at The preferred submission format is pdf.

When writing the final paper you should take into account all of the reviewers' feedback and suggestions. In addition, all papers, in particular those that are more theoretical, must clearly establish the relevance of the work to health care.

Please note the final version of the paper will be reviewed by workshop chairs to assure the high quality standard. A failure to address any of the above points may result in your paper not being published on the workshop website before the workshop.

Instructions for the workshop presenters

Call for papers

We seek paper submissions developing new or applying existing ML methods to medical and health-care applications. The topics of interest include, but are not limited to:  Submissions addressing theoretical problems should clearly outline the expected impact of the proposed solution to the medical field.

Paper submission

Please submit an extended abstract (1 to 3 pages in two-column ICML format) to the workshop email address
The abstract should include author names, affiliations, and contact information. Papers will be reviewed by the members of the program committee and decisions on the acceptance together with the reviewers' feedback will be emailed back to authors on May 17, 2008. Authors of accepted extended abstracts are encouraged to submit a full version of their paper. All submissions will be published on the conference web site.


In addition to authors of accepted papers presenting their works we are planning on having 4-5 invited talks.

Past Important Dates


Program Committee

Upcoming Related Events

If you know of any other events please send us an e-mail at