Progress in characterizing and imaging prostate tissues for guiding biopsies and planning and targeting treatment of prostate cancer

Our objective is to develop imaging methods that can distinguish viable cancerous tissue from other prostate tissues in order to improve biopsy guidance and treatment targeting. To do this, we acquired ultrasonic RF echo-signal data and clinical variables, e.g., PSA, for over 3,000 biopsies, and computed spectral-parameter values for each biopsied region. Using a neural network trained with these data, we generated a lookup table that translated parameter values at each pixel location into a local score for cancer likelihood. Images displaying the map of local scores are called tissue-type images (TTIs). ROC-curve areas were greater for neural-network classification than for classification using conventional B-mode-based methods, and predicted a sensitivity improvement of more than 20% over conventional, ultrasound-guided biopsies. We are initiating studies of prostate tissue characterization using magnetic resonance spectroscopy (MRS), and are investigating the feasibility of combining information from ultrasound and MRS for improved imaging of prostate cancer. TTIs applied in real time may markedly improve cancer-detection by directing biopsies to cancerous regions. TTI data combined with MRS parameters potentially can provide a powerful new hybrid 3-D imaging method for detecting, evaluating, and treating prostate cancer.