Project Name: Visualization of nearest-neighbors

Sponsor:
William C. Shoemaker (email: wcshoemaker00@hotmail.com)
Roger Jelliffe (email: jelliffe@usc.edu)

Partners:
None

Background:
Researches predict a patient’s probability of survival using the Survival Prediction Model (SPM). The concept of nearest-neighbor is based on SPM. Clinicians decide the best course of action for a new patient from previous patient data (characteristics and outcomes) with similar characteristics. The Therapeutic Decision Support System (TDSS) is SPM’s practical counterpart. TDSS provides real-time guidance for optimizing drug therapy.

Problem:
Visualization of the Survival Prediction Model output will help clinicians decide whether more or fewer neighbors (e.g. patients) are needed to get a more accurate prediction. Currently, the output is presented as an n-dimensional array of numbers. Visualization of the data can help analyze whether the Survival Prediction Model arbitrary selection of a number is too lose or too tight.

Constraints:
Use of COTS tool for data visualization. The current tool’s output will be used as input to the new tool being developed. The proposed system should be compatible with Windows 2000 and up.

Desired Deliverables:
A COTS tool that graphically displays the output of the Survival Prediction Model in terms of different standard deviations is desired. LAPK already uses the COTS products, Stingray, by Dazdaq and ProEssentials, by Gigasoft, therefore they should be considered before any others.