COMPUTER-AIDED DIAGNOSIS OF GASTRIC CANCER

35,529 cases examined fibergastroscopically are collected from 52 domestic papers for calculation of conditional and prior probability values of 4 classes of diagnostic results. According to the program, these values are taken as computer input. When a new patient is diagnosed, the output is in the form of probability values of various suspected disorders. The maximum probability value is taken as the result of computeraided diagnosis (CAD). The CAD accuracy of 58 gastric cancer cases confirmed pathologically is 87.93% and reliability 89.47%. Total agreement rate of 132 cases in 4 classes is 81.82%. CAD is promising in differential diagnosis of gastric cancer. Statistical analysis shows that the accuracy and reliability of CAD are superior to barium gastroenterography and experienced physician diagnosis.