Artificial neural network analysis (ANNA) of prostatic transrectal ultrasound.

Background: Our purpose was to determine the diagnostic potential of a new, computerized method of interpreting transrectal ultrasound (TRUS) information by artificial neural network analysis (ANNA). This method was developed to resolve the current dilemma of visual differentiation between benign and malignant tissue on TRUS. To train and objectively evaluate ANNA, a new precise method of computerized virtual correlation of preoperative ultrasound findings and radical prostatectomy histopathology was devised. After training with this pathologically confirmed digitized TRUS information, ANNA was tested in a blinded study.

Methods: Following radical prostatectomy, 289 pathology whole-mount sections of 61 patients were correlated digitally with the corresponding TRUS slices. Specific selection of TRUS areas unequivocally identified on the correlated digitized pathohistology resulted in 553 pathology-confirmed representations (samples). Of these, 53 were used for training and 500 were subjected to blind analysis by ANNA.

Results: ANNA classified 378 (99%) of the 381 benign pathology-confirmed samples correctly as benign. The false-positive rate was 1% (n = 3). Of the 119 pathology-confirmed malignant samples, 94 (79%) were classified correctly; 25 (21%) were falsely classified as normal. Out of all 119 cancers, ANNA classified 60 (71%) of the hypoechoic cancers as malignant and 24 (29%) as benign. Surprisingly, 34 (97%) of the isoechoic cancers were correctly classified by ANNA, missing only one sample.

Conclusions: The introduction of ANNA enhanced the accuracy of TRUS prostate cancer identification. Although not all malignant areas were detected, cancer was detected in each patient. The ability to detect isoechoic cancerous lesions appears to be the essential innovation over conventional TRUS interpretation.