Start date: 23-02-2022
End date: 23-11-2022
In prostate disease, biopsies, small tissue specimens, play a prominent role in the early detection and staging of cancer. Radical prostatectomy, complete surgical removal of the prostate, is typically performed as a curative treatment following a biopsy positive for prostate cancer. Radical prostatectomies result in enormous amounts of tissue for the pathologist to analyse.
Prostatectomy specimens need to be fully graded to allow determination of an adequate follow-up strategy for the patient. For example, a small high-grade component of cancer might be present that alters the treatment strategy and was missed on initial biopsy. Our group has developed a prostate cancer grading algorithm for biopsy samples. Next, we plan to extend this algorithm to prostatectomy data. From an algorithmic perspective, analysis of prostatectomy specimens is challenging due to the presence of a wide variation of tissue, some of which only appear in a fraction of cases. Examples include the seminal vesicles, surrounding fat, nerves, muscle, and large blood vessels, among others. These tissue types are also rarely sampled in biopsies, such that our biopsy-trained algorithm fails on those cases.
We will use active learning to iteratively enrich the training dataset with normal tissue from various locations, depending on which of them throws off the algorithm the most. In order to ensure training data contains adequate tissue variability, we will use ‘spatial hard-negative mining’ to monitor misclassifications in normal tissue during training time in a location-aware manner and sample data accordingly.
In this project we plan to combine data from public datasets PANDA (~11.000 annotated slides) and PESO, and inhouse dataset of prostatectomy data (slides from ~40 patients with detailed annotations + ~4000 slides with only slide level labels).
The algorithm will be made publically available as a Docker container on https://grand-challenge.org/.