SKIN CANCER RECOGNITION BY COMPUTER VISION

Period of Performance: 01/01/1987 - 12/31/1987

$244K

Phase 2 SBIR

Recipient Firm

William V. Stoecker And
1100 West 10th
Rolla, MO 65401
Principal Investigator

Abstract

A FEASIBILITY STUDY IS PROPOSED FOR A COMPUTER VISION DEVICE TO HELP GENERAL PRACTITIONERS IN DIAGNOSING SKIN CANCERS. THE OVERALL GOAL IS TO COMBINE A CURRENTLY AVAILABLE MEDICAL EXPERT SYSTEM COMPUTER PROGRAM WITH A COMPUTER VISION FRONT END TO CREATE A USEFUL DIAGNOSTIC TOOL FOR SKIN CANCER. THE PROPOSED PHASE I RESEARCH WILL DEVELOP MODEL ALGORITHMS FOR A KNOWLEDGE BASED FEATURE DETECTION SYSTEM FOR 35 MM PHOTOGRAPHIC SLIDES OF SKIN TUMORS. THE RESEARCH WILL USE A VAX 11/780 COMPUTER AND A GOULD DEANZA IP-8400 IMAGE PROCESSING SYSTEM. WE HOPE TO DETECT THE TUMOR FEATURES WITH HIGHEST TUMOR PREDICTIVITY AND TO OUTLINE APPROACHES TO DETECT THE REMAINING TUMOR FEATURES DURING PHASE I. IF THIS APPROACH FOR SKIN CANCER DETECTION IS SUCCESSFUL, IT HAS THE POTENTIAL TO SIGNIFICANTLY REDUCE THE NATION'S ANNUAL COST FOR SKIN TUMOR DETECTION AND TREATMENT, CURRENTLY IN EXCESS OF ONE BILLION DOLLARS. EARLIER CANCER DETECTION ALSO HAS THE POTENTIAL OF RELIEVING UNTOLD MORBIDITY AND ELIMINATING UNNECESSARY BIOPSIES.