Expert system for automatic phono-mechanocardiographic diagnosis

Abstract

An expert system has been created for collecting, controlling, and categorizing phonomechanocardiographic data. LV (left ventricle) contraction, ejection, relaxation, and ventricular filling is detected by means of an apexcardiogram. The various forms and combinations of valvular heart disease are characterized by the phonospectra of the auscultation areas. The expert system Mechano has rule-based knowledge representation, backward chaining, uncertainty handling and explanation facilities. The database consists of data on 200 patients with ischemic heart disease and 140 patients with valvular heart disease. By using question nets the digitalized knowledge elements of the apex (amplitude-time intervals) and the phonomap (location, duration, amplitude of sounds, and murmurs) are automatically collected. The diagnosis of N different forms of valvular heart disease and the evaluation of LV functions are based on diagnosis rules containing values of possibilities determined by multiple statistical models (regression, factor, and cluster). The program, implemented on a PC, is written in Turbo Prolog.<>