Team: 43
School: LAS CRUCES HIGH
Area of Science: Radiology - Medicine
Abstract:
Automated Analysis of Bone Scans
Problem:
"Hot Spots" are areas in a bone scan that suggest areas that might contain cancerous material. Hot Spots are indicated by an area of above average brightness in a bone scan. Radiologists scan a bone scan for Hot Spots by eye, subjecting a diagnosis to human error such as not seeing a Hot Spot. In the case in which of two or more Hot Spots' brightness is compared, Radiologists must use their own judgment, subjecting a diagnosis to human error. Reading bone scans is a time consuming process. When Radiologists adjust the brightness/contrast of a bone scan, faux Hot Spots are occasionally created.
Solution:
Devise a program that scans a bone scan pixel by pixel for Hot Spot areas. This eliminates the need for Radiologists to ever have to search a bone scan for Hot Spots. Each pixel of the bone scan will be given a numerical value to indicate brightness. Thus, when comparing Hot Spots, Radiologists will be able to compare solid numerical values of specific Hot Spot datum, eliminating the estimation required to find the brighter of two or more Hot Spots. Using a computer to scan a bone scan will result in a quick bone scan. Users will be able to define brightness value ranges that are to be marked by the program as Hot Spots. This threshold will increase/decrease proportionally to any brightness/contrast changes of the image made by the Radiologist, thus eliminating faux Hot Spots.
Procedure:
Program will be written in Java and exported as a Windows executable. Program will contain two parts, as follows. In the automated part of the program, the program will scan every pixel of a bone scan and define the RGB values of each pixel. It will then compare these RGB values to a user defined threshold and then mark each pixel as a Hot Spot or not, hopefully using an image overlay. An alternate solution to automated search would be the creation of a contrast filter.
When contrast is increased, threshold values also increase proportionally. In the manual part of the program, users will be able to find specific values of pixel color data with their cursors. Pixels over a user-defined threshold will output with a notification tag. Pixel data will be displayed with information including x-y coordinates, RGB values of that pixel (1-255, 255 being the brightest) and a percentage value of how bright the pixel is compared to black (white = 100%, black = 0%)
Team Members:
Julia Silva
daniel parrott
alan hshieh
Jerry Yeh
Sponsoring Teacher: Gregory Marez