Artificial Intelligence: A Modern Approach
Artificial Intelligence: A Modern Approach
Inference in the Promedas Medical Expert System
AIME '07 Proceedings of the 11th conference on Artificial Intelligence in Medicine
| Hi-index | 0.00 |
The possibility of developing systems to support medical diagnosis through Artificial Intelligence (AI) allows conceiving Expert Systems (ES), which constitute successful methods to solve AI problems to program intelligent systems. This work deals with creating an ES to support the diagnosis of cervical lesions by identifying them in colposcopic images; for this purpose, 140 images were analyzed, with the most interesting and relevant result from this action being the definition of discriminating features: surface, color, texture and edges. These features will be used to evaluate an image and offer diagnosis as established by the expert physician, like: normal, inflammation process, immature metaplasia, gland eversion and low-grade lesion. To evaluate the system's performance, we obtained support from an expert colposcopy physician, who evaluated all 140 images. The results indicated that the ES obtained an efficiency of 75.75 % and an error percentage of 20.405%, including 4.04% that was not evaluated by the expert, who declared that the region or lesion was impossible to identify because the image was not clear.