Full-field optical coherence tomography (FFOCT) offers a fast and non-destructive method of obtaining images of biological tissues at ultrahigh resolution, approaching traditional histological sections. In the context of prostate cancer diagnosis involving multiple biopsies, FFOCT could be used to validate the cores just after they are obtained in order to guide the number of biopsies to be performed. The aim of the study was to define and test a training protocol for efficient FFOCT prostate biopsy assessment. Three readers (a pathologist with previous experience with FFOCT, a pathologist new to FFOCT, and a urologist new to FFOCT) were trained to read FFOCT images of prostate biopsies on a set of 20 commented zooms (1 mm field of view) and 25 complete images. They were later tested on a set of 115 anonymized and randomized images of prostate biopsies. The results showed that an extra 30 images were necessary for more complete training as compared to prior studies. After training, pathologists obtained 100% sensitivity on high-grade cancer detection and 96% overall specificity; the urologist obtained 88% sensitivity on high-grade cancer and 89% overall specificity. Overall, the readers obtained a mean of 93% accuracy of qualifying malignancy on prostate biopsies. Moreover, the two pathologists showed a steeper learning curve than the urologist. This study demonstrates that a training protocol for such a new imaging modality may be implemented and yield very high efficiency for the prehistologic detection of malignancy on prostate biopsies.
原文链接
分享
引用
