Computer-aided image analysis in transrectal ultrasound of the prostate

Summary

To objectify the interpretation of prostate, ultrasound images were processed by a computer-based image analysis system (IAS). The IAS “tissue descriptors” are not dependent on the gray scale. Transrectal ultrasound (TRUS) images, results of the IAS and pathologic whole mounts (PWM), were correlated in an attempt to define the efficiency of the IAS in differentiating carcinoma from normal tissue of the prostate. Using the closely correlated TRUS and PWM slices, a restrictive setting (high specificity) of the IAS yielded rates of 90% (true positive), 10% (false negative) and 5% (false positive). Using a less restrictive setting (higher sensitivity), rates of 100% (true positive), 10% (false negative) and 12% (false positive) were noted. These encouraging results were obtained from the peripheral zone of the prostate. However, the clinical false-positive and false-negative rates for IAS have not yet been determined.