[The prototype of an expert system for the diagnosis of pseudotumorous lesions and tumors of the jaws: ADAPT-M. Archiviazione e Diagnosi Automatica di Pseudotumori e Tumori delle ossa Mascellari.]

Many different tumors and pseudotumors may affect the jaw bones. The number and the rarity of most of these lesions, however, make both classification and differential diagnosis quite difficult. Computer software for statistical calculations and logical-deductive reasoning on vast amounts of data can improve diagnostic skills. These applications are defined as computer-assisted decision-making, medical decision support, or expert systems. This study was aimed at developing a prototype probabilistic expert system, based upon knowledge from an 'ad hoc'computerized data base, as an aid in the radiologic diagnosis of jaw tumors and pseudotumors. This program has been called ADAPT-M. The study considered 92 patients with benign space-occupying and fully documented lesions of the jaw bones. For each case, a list of parameters concerning different radiologic exams was considered. From all these pieces of information a data base was built, to calculate both the prevalence of each type of lesion and the frequency of many variables in the single conditions. For each kind of lesion 44 variables were considered. ADAPT-M used a formula based on Bayes' theorem to calculate the 'a posteriori' probability of a diagnosis in the presence of a symptom. Overall diagnostic probability rate was high when the highest score hypothesis was matched with pathologic findings (80%) and even higher (96.1%) when the two most probable diagnoses were considered together. As expected, ADAPT-M had higher sensitivity when used with lesions with typical semiology. This results in an unquestionable limitation, especially in the patients in whom a predictive diagnosis would be most desirable. The creation of a larger data base of known cases and software development may help increase the diagnostic accuracy of the ADAPT-M system.