Improved prediction of prostate cancer recurrence based on an automated tissue image analysis system

摘要

Prostate tissue characteristics play an important role in predicting the recurrence of prostate cancer. Currently, experienced pathologists manually grade these prostate tissues using the GIeason scoring system, a subjective approach which summarizes the overall progression and aggressiveness of the cancer. Using advanced image processing techniques, Aureon Biosciences Corporation has developed a proprietary image analysis system (MAGIC/spl trade/), which here is specifically applied to prostate tissue analysis and designed to be capable of processing a single prostate tissue hematoxylin-and-eosin (H&E) stained image and automatically extracting a variety of raw measurements (spectral, shape, etc.) of histopathological objects along with spatial relationships amongst them. In the context of predicting prostate cancer recurrence, the performance of the image features is comparable to that achieved using the GIeason scoring system. Moreover, an improved prediction rate is observed by combining the GIeason scores with the image features obtained using MAGIC/spl trade/, suggesting that the image data itself may possess information complementary to that of GIeason scores.