Simon K. B. Spohn

Purpose: Prostate cancer (PCa) is the most frequently diagnosed malignancy among men in Germany. Advances in diagnostics and treatment have transformed PCa into a chronic disease. Given the heterogeneity of PCa, there is a need for additional stratification tools. This review focuses on updating the evidence for genomic classifiers (GC; Decipher [Veracyte Inc. San Diego, CA, USA], Prolaris [Myriad Genetics, Inc., Salt Lake City, UT], and Oncotype DX [Exact Sciences, Madison, WI, USA] tests) and artificial intelligence (AI)-based digital histopathology biomarkers (ArteraAI Prostate Test) in the context of radiotherapy (RT) for PCa. Methods: The members of the Prostate Cancer Working Group of the German Society of Radiation Oncology (DEGRO) conducted an updated literature search on GCs and histopathological biomarkers in PCa, covering original articles published between January 2022 and February 2024 in the PubMed database. Results: In addition to previous reviews, 11 relevant studies were identified, of which nine studies analyzed biomarkers within prospective phase II or III trials. Eight trials focused on genomic biomarkers, of which three addressed GCs in primary localized PCa, three in recurrent PCa in the setting of salvage RT, and two in metastatic castration-sensitive PCa. In localized PCa, GCs could be validated in a retrospective analysis of randomized controlled trials. Additionally, three studies reported on AI-based histopathology biomarkers. Conclusion: Genomic classifiers and AI-based digital histopathology models might have superior prognostic and predictive value compared to established clinical and pathological parameters in localized, recurrent, and metastatic PCa. Despite promising results, prospective validation of these biomarkers in randomized trials remains limited. This review underscores the need for further prospective trials to confirm the usefulness of these biomarkers in PCa.

Accurate segmentation of prostate tumours from PET images presents a formidable challenge in medical image analysis. Despite considerable work and improvement in delineating organs from CT and MR modalities, the existing standards do not transfer well and produce quality results in PET related tasks. Particularly, contemporary methods fail to accurately consider the intensity-based scaling applied by the physicians during manual annotation of tumour contours. In this paper, we observe that the prostate-localised uptake threshold ranges are beneficial for suppressing outliers. Therefore, we utilize the intensity threshold values, to implement a new custom-feature-clipping normalisation technique. We evaluate multiple, established U-Net variants under different normalisation schemes, using the nnU-Net framework. All models were trained and tested on multiple datasets, obtained with two radioactive tracers: [68-Ga]Ga-PSMA-11 and [18-F]PSMA-1007. Our results show that the U-Net models achieve much better performance when the PET scans are preprocessed with our novel clipping technique.