Computerized classification of corpus cavernosum electromyogram signals by the use of discriminant analysis and artificial neural networks to support diagnosis of erectile dysfunction.

Corpus cavernosum electromyogram (CC-EMG) provides diagnostic information on cavernous autonomic innervation and a measure of the degree to which the cavernous smooth muscle cells are intact. The complicated CC-EMG is evaluated and used in the diagnosis of patients suffering from erectile dysfunction. The evaluation procedure has been simplified by applying digital signal processing techniques. Since mathematically-based interpretations require quantitative data, spectral analysis was performed. The derived biosignals were analyzed by fast Fourier transform (FFT). Besides various other spectral parameters, specific frequency bands were determined in the power spectrum using factor analysis. The parameters were used for the computerized classification of normal and pathological CC-EMG data and the classification was performed using two independent methods: discriminant analysis (DA) and artificial neural networks (ANN). A medical expert analyzed a total of 200 CC-EMG recordings from patients with and without erectile dysfunction and separated these into normal (136) and pathological (64) cases. Although each independent method had already resulted in a relatively high number of correct classifications, the classification success rate could be slightly improved by using a combination of both classification methods. A total of 72.79% and 77.94% were successfully classified using DA and ANN, respectively. The combination of both methods increased the classification success to 80.15%. The results of this study enabled impartial evaluation of the CC-EMG signals for clinical diagnostic purposes of erectile dysfunction. This method provided an objective and easy way to analyze the CC-EMG. Furthermore, this results in patient diagnosis becoming an easier task for less experienced doctors, since little knowledge of the raw signal is needed.