The mission of the Scientific Computing Section is to support and promote the goals of DCRI by seeking and engaging in active research and collaborative work within the NIH community and with academia and industry. This work is intended to advance knowledge in the biomedical and computing sciences, and to produce results that lead to improved health for everyone.

Full Time Staff: Jim DeLeo, Section Chief, jdeleo@nih.gov, 301-496-3848; Carl Leonard, Engineer and Applications, cleonard@lired.com, 301-496-0191

Student Interns: Simone Campbell, Gaithersburg High School, Morgan Clinton, University of South Carolina, Rishi Gharpuray, Stanford University, James Woo, Johns Hopkins University

Collaborators: Ashok Asthagiri, Mark Green, Fred Miller, Sachi Rath, Alan Remaley, Monika Skarulis

Supporters: Aaron Baughman, Judy Dayhoff, Alexandru Floares, Adam Gaweda, Jerry McLaughlin, Nancy Miller, Chuck Selden, George Thoma and Lotfi Zadeh

What We Do: The Scientific Computing Section develops and applies medical analytics [disclaimer] computational methodologies, i.e. methodologies that use data to discover new knowledge [disclaimer] and to support decision making [disclaimer]. In medicine much knowledge and decision making is tacit or implicit [disclaimer], meaning it is based on past experiences, rules of thumb and general qualitative assessments. The term medical analytics [disclaimer] as used here subsumes and includes statistics [disclaimer], informatics [disclaimer], data mining [disclaimer], intelligent computing [disclaimer], soft computing [disclaimer] and evolutionary computing [disclaimer]. By applying such tools we attempt to expose implicit [disclaimer] knowledge and express it in a more explicit [disclaimer]formulation and therefore make it more readily transferable and more widely usable. Some of our major activities in 2008 are described next.

Computational Methodology Development: The following computational methodologies were developed this year:

HubuH^TM: HubuH^TM [disclaimer] is a new, simple yet powerful clustering algorithm [disclaimer]. It groups patients and events into meaningful clusters according to their feature values [disclaimer]. In 2009 we will expand HubuH^TM to identify significant clusters and to automatically generate written narrative and graphical reports that describe the members of the significant clusters by describing characteristics of their feature values [disclaimer].

Medical Composite Index: The medical composite index [disclaimer] is a classification algorithm [disclaimer] that accounts for prevalence [disclaimer] and misclassification costs [disclaimer]. It can be applied to any bivalent [disclaimer] or ordinal scale [disclaimer] outcome problem such as the presence of a disease, adverse drug reaction, biopsy outcome, survival up to a specified time point, or an ordinal scale [disclaimer] value like the Gleason Score [disclaimer] as used in prostate cancer diagnosis.

I-Biopsy: The “i-biopsy” [disclaimer] is a particular instance of the medical composite index [disclaimer] concept just described. With it we are attempting to predict the outcome of a biopsy without the biopsy actually being performed. This is done by building a medical composite index [disclaimer] system that utilizes multiple patient feature and marker values to predict the outcome. An NIH Employee Invention Report describing the “i-biopsy” [disclaimer] concept was filed in November, 2008.

Intelligent Assay Design: An assay [disclaimer] is a procedure in which properties or concentrations of analytes [disclaimer] are measured. In 2009 we may explore the possibility of developing intelligent software to facilitate custom designing efficient assays to target diagnoses of specific diseases.

Data Farming: Often projects can be delayed because data sets are sparse or not yet available. We have begun developing an approach to data farming [disclaimer] which means “growing” data. This methodology is ad hoc, highly nonparametric [disclaimer], and can be subjected to validation checks. When course features of anticipated data are known, “grown” data samples can be generated with this methodology so that medical analytics [disclaimer] software development can proceed while real data is being generated and prepared.

Data Fusion: Data fusion [disclaimer] refers to computer science [disclaimer] techniques that combine data from multiple disparate sources in order to achieve more efficient and accurate inferences than achievable from a single data source. Since our biomedical collaborators usually have data from disparate sources, we have begun exploring known data fusion techniques that could be useful and we will continue to do so in 2009 and use them where appropriate.

Collaborations: We continued and started clinical research collaborations in which medical analytics [disclaimer] methodologies such as those mentioned above are developed and applied. Medical subjects in these collaborations include inflammatory diseases, cholesterol formation, obesity research, addictive behaviors, brain and nervous system tumors, cervical cancer, prostate cancer and biopsy replacement technology.

SMART-MART^TM: In 2008, we started developing SMART-MART^TM [disclaimer] – a computer server that houses our collaborators’ data with appropriate medical analytics software such as that developed by the Scientific Computing Section.

BTRIS Link: BTRIS is the NIH Clinical Center Biomedical Translational Research Information System. We built a link between BTRIS and SMART-MART^TM [disclaimer] and hope to make SMART-MART^TM [disclaimer] available to BTRIS users as SMART-MART^TM resources develop.

Summer Students: We had another highly successful summer student program [disclaimer] in 2008; four students, four BCIG lectures, four posters and one peer-reviewed published paper.

Behavioral Health Measurements: Behavioral health [disclaimer] refers to the relationship between human behavior and overall health. In September 2008 the International Conference on Outcomes Measurement: Focus on Behavioral Health [disclaimer] hosted by NIDA and the Office of Behavioral Social Science Research (OBSSR) [disclaimer] was held in Bethesda, Maryland at which it was indicated that the need for improved measurement in health care and other fields has increased dramatically demands for better measurement to enable better management. Inspired by this plea, we developed an approach to quantitatively and graphically present complex constructs such as depression, addiction, criminality and intelligence into a set of less abstract orthogonal terms as a methodology to compliment or replace forced linear scales currently in use. In 2009, we hope to continue with this work with an NIH collaborator.

Marvin Minsky: Professor Marvin Minsky [disclaimer] is considered to be the father of artificial intelligence (AI) [disclaimer]. We are working with members of the NIH Office of the Director, Lister Hill Center, and NIDDK to bring Professor Marvin Minsky [disclaimer] to the NIH in 2009 to deliver an NIH Director’s Lecture, to speak at a BCIG [disclaimer] special event, and to meet with him in small dialogue groups to explore possible specific roles for modern AI in medicine [disclaimer].

Biomedical Computing Interest Group (BCIG): We managed another highly successful (BCIG) program in 2008 as indicated by the following characteristics:

• 17 events
• 850 members
• abridged announcements with web links to more details
• most events webcast and archived for easy access
• actively accessed BCIG web site [disclaimer]
• web site had 25,000 visits in 2008 (70 per day)
• new web-based monthly BCIG News Letters [disclaimer]
• expanded participation in the BCIG Book Club [disclaimer]
• BCIG past events library [disclaimer]
• 2009 program 60% planned

Resource Seeking: At the end of the year more active seeking for collaborators and mechanisms to bring in resources to expand the Scientific Computing Section were initiated.

NIH Intramural Scientific Computing Initiative: Late in the year we began exploring the possibility of creating an NIH-wide requirements, resource and idea sharing initiative to use medical analytics [disclaimer] more effectively and efficiently across the NIH mission.

Publications

Clinton, Morgan, DeLeo, Jim, Grafman, Jordan. Corticobasal Syndrome Patient Subtyping with HubuH^TM – a New Clustering Algorithm. October 2008. Fifth International Meeting on Computational Intelligence Methods for Bioinformatics and Biostatistics, IIASS, Vietri sul Mare, Salerno, Italy

DeLeo, Jim. BCIG Monthly Newsletters published on the NIH Biomedical Computing Interest Group (BCIG) Web Site.

DeLeo, Jim & Leonard, Carl. Maintaining BCIG web site which includes text and graphics design, writing and posting announcements and video archiving programs.

Presentations

DeLeo, Jim & Leonard, Carl. February 28, 2008. Detailed reviews of selected chapters of “I Am a Strange Loop,” by Douglas Hofstadter for the NIH Biomedical Computing Interest Group (BCIG). (These presentations are viewable on the BCIG web site.)

DeLeo, Jim & Leonard, Carl, March 27, 2008. Detailed reviews of selected chapters of “Introducing Artificial Intelligence,” by Henry Brighton and Howard Selina for the NIH Biomedical Computing Interest Group (BCIG). (These presentations are viewable on the BCIG web site.)

DeLeo, Jim & Leonard, Carl, April 24, 2008. Detailed reviews of selected chapters of “Invented Reality: How Do We Know What We Believe We Know,” by editor Paul Watzlawick for the NIH Biomedical Computing Interest Group (BCIG). (These presentations are viewable on the BCIG web site.)

DeLeo, Jim. July 2008. Scientific Computing Opportunities at the NIH. This was an invited ad hoc presentation given at the Evolutionary Computation in Practice "Ask the Experts" track at the 2008 Genetic and Evolutionary Computing Conference (GECCO 2008)

Woo, James, August 6, 2008. SMART-MART^TM - a Desktop Tool for Clinical Data Mining. This was an NIH Biomedical Computing Interest Group (BCIG) presentation. James Woo is a student at Johns Hopkins University. DCRI Mentors: Jim DeLeo & Carl Leonard.

Gharpuray, Rishi, August 6, 2008. Facilitating Data Analysis through Ruby & R Interaction. This was an NIH Biomedical Computing Interest Group (BCIG) presentation. Rishi Gharpuray is a student at Stanford University. DCRI Mentors: Jim DeLeo & Carl Leonard.

Campbell, Simone, August 6, 2008. Understanding Behavior and Brain Activity Using fMRI Imaging. This was an NIH Biomedical Computing Interest Group (BCIG) presentation. Simone Campbell is a high school student at Gaithersburg High School in Gaithersburg Maryland. DCRI Mentors: Jim DeLeo & Carl Leonard.

Clinton, Morgan, August 6, 2008. Artificial Neural Network Subgrouping of Corticobasal Syndrome Patients. This was an NIH Biomedical Computing Interest Group (BCIG) presentation. . Morgan Clinton is a student at the University of South Carolina. DCRI Mentors: Jim DeLeo & Carl Leonard.

DeLeo, Jim. September 25, 2008. A detailed review of selected chapters of “Genesis Machines: The New Science of Biocomputing,” by Martyn Amos for the NIH Biomedical Computing Interest Group (BCIG). (This review is viewable on the BCIG web site.)

DeLeo, Jim. September 28, 2008. “The Human Use of Human Beings: Cybernetics and Society,” by Norbert Weiner. This included reviews of The Idea of a Contingent Universe (Preface), Cybernetics in History (Ch. 1), The Mechanism and History of Language (Ch. 4), and Organization as Message (Ch. 5) presented at a meeting of the Washington Evolutionary Systems Society (WESS) at Georgetown University.

Clinton, Morgan, DeLeo, Jim, Grafman, Jordan. October 2, 2009 Corticobasal Syndrome Patient Subtyping with HubuH^TM – a New Clustering Algorithm. Fifth International Meeting on Computational Intelligence Methods for Bioinformatics and Biostatistics, IIASS - Vietri sul Mare, Salerno (Italy). (presented by Jim DeLeo)

Dayhoff, Judy & DeLeo, Jim. October 2, 2008 The Intelligent Coin: A Statistical Test for Pattern Classification in Medicine and Bioinformatics. Fifth International Meeting on Computational Intelligence Methods for Bioinformatics and Biostatistics, IIASS - Vietri sul Mare, Salerno (Italy), (presented by Jim DeLeo)

Turano, Augie & DeLeo, Jim. October 2. Protocol Adherence versus Pattern Recognition in Healthcare Informatics. Fifth International Meeting on Computational Intelligence Methods for Bioinformatics and Biostatistics, IIASS - Vietri sul Mare, Salerno (Italy), October 3, 2008 (presented by Jim DeLeo)

DeLeo, Jim & Leonard, Carl. October 23, 2008. Detailed reviews of selected chapters of “The Emotion Machine: Commonsense Thinking, Artificial Intelligence, and the Future of the Human Mind,” by Marvin Minsky for the NIH Biomedical Computing Interest Group (BCIG). (These presentations are viewable on the BCIG web site.)

DeLeo, Jim. November 22, 2008. “The Synaptic Self: How Our Brains Become Who We Are,” by Joseph LeDoux. This included reviews of Building the Brain (Ch.4) and Adventures in Time (Ch. 5) at a meeting of the Washington Evolutionary Systems Society (WESS) at Georgetown University.

DeLeo, Jim. December 1, 2008. Biomedical Data Mining. A presentation requested by the Acting Director of NIH Intramural Obesity Research Initiative.

Posters

Woo, James, Leonard, Carl & DeLeo, Jim. 2008. SMART-MART^TM - a Desktop Tool for Clinical Data Mining. NIH Summer Research Poster Session. James Woo is a student at Johns Hopkins University. DCRI Mentors: Jim DeLeo & Carl Leonard.

Gharpuray, Rishi, Leonard, Carl & DeLeo, Jim. 2008. Facilitating Data Analysis through Ruby & R Interaction. NIH Summer Research Poster Session. Rishi Gharpuray is a student at Stanford University. DCRI Mentors: Jim DeLeo & Carl Leonard.

Campbell, Simone, Leonard, Carl & DeLeo, Jim. 2008. Understanding Behavior and Brain Activity Using fMRI Imaging NIH Summer Research Poster Session. Simone Campbell is a high school student at Gaithersburg High School in Gaithersburg Maryland. DCRI Mentors: Jim DeLeo & Carl Leonard.

Clinton, Morgan, Leonard, Carl, DeLeo, Jim. 2008. Artificial Neural Network Subgrouping of Corticobasal Syndrome Patients

NIH Summer Research Poster Session. Morgan Clinton is a student at the University of South Carolina. DCRI Mentors: Jim DeLeo & Carl Leonard.

Publications, Presentations, Special Sessions & Workshops for 2009

DeLeo, Jim. Biomedical Data Mining. Early 2009. A lecture requested by the Chief of the Clinical Genetics Branch, NCI.

DeLeo, Jim. Introduction to Evolutionary Computing. Early 2009. A lecture requested by Lister Hill Center OD, NLM.

DeLeo, Jim. Data Acquisition and Scientific Computing in Clinical, Translational and Basic Science Research. March 10, 2009. An invited lecture to be given at a NIH Surgical Neurology Branch Science Conference.

Leonard, Carl & Berg-Cross, Gary. March 2009. Developmental Robotics. NIH Biomedical Computing Interest Group (BCIG) lecture.

DeLeo, Jim, Leonard, Carl & Miller, Fredrick, 2009. An Artificial Neural Network Approach to Identifying Idiopathic Inflammatory Myopathy Patients.

(a computer science journal TBD)

Miller, Fredrick, Leonard, Carl & DeLeo, Jim, 2009. An Artificial Neural Network Approach to Identifying Idiopathic Inflammatory Myopathy Patients.

(a medical journal TBD)

DeLeo, Jim. June 2009. Computational Intelligence in Direct Support of Patient Care – a proposed and accepted special session for the 2009 International Joint Conference on Neural Networks (IJCNN), Atlanta, Georgia.

DeLeo, Jim. July, 2009. Genetic and Evolutionary Computing in Support of Patient Care – a proposed workshop for the 2009 Genetic and Evolutionary Computing Conference (GECCO), Montreal, Canada.

DeLeo, Jim. June 2009. The Medical Composite Index An Artificial Neural Network Approach to Classification That Accounts for Prevalence and Misclassification Costs in Predicting Medical Outcomes” – a proposed paper and presentation to be presented at the International Joint Neural network Conference (IJCNN) 2009 in Atlanta Georgia.

DeLeo, Jim & Thoma, George. July 2009. How Can Evolutionary Computing Help Us Monitor Electronic Patient Records to Suggest Patient Management Options? – a proposed paper and presentation to be presented at the Genetic and Evolutionary Computing Conference (GECCO) 2009 in Montreal, Canada.

DeLeo, Jim. July 2009. Practical Evolutionary Computing Opportunities in Medicine. This will be an invited talk in the Evolutionary Computation in Practice "Ask the Experts" track at the Genetic and Evolutionary Computing Conference (GECCO) 2009 in Montreal, Canada.

DeLeo, Jim & Leonard, Carl. 2009. We are planning 15 or more Biomedical Computing Interest Group (BCIG) events including lectures, book club meetings, brainstorming sessions and tutorials for 2009.