Renal cell carcinoma is the third type of urologic cancer and has a poor prognosis with 30% of metastatic patients at diagnosis. The antiangiogenics and targeted immunotherapies led to treatment remodeling emphasizing the role of the tumour microenvironment. However, long-term responses are rare with a high rate of resistance. New strategies are emerging to improve the efficacy and the emerging drugs are under evaluation in ongoing trials. With the different treatment options, there is an urgent need to identify biomarkers in order to predict the efficacy of drugs and to better stratify patients. Owing to the limitations of programmed death-ligand 1 (PD-L1), the most studied immunohistochemistry biomarkers, and of the tumor mutational burden, the identification of more reliable markers is an unmet need. New technologies could help in this purpose.

Renal cell carcinoma encompass distinct diseases with different pathologic features and distinct molecular pathways. Immune checkpoint inhibitors targeting the programmed death receptor ligand 1 (PD-L1)/programmed death receptor 1 (PD-1) pathway alone or in combination have greatly changed clinical management of metastatic renal cell carcinoma, now competing with antiangiogenic drugs in monotherapy for first-line treatment. However, long-term response rates are low, and biomarkers are needed to predict treatment response. Quantification of PD-L1 expression by immunohistochemistry was developed as a promising biomarker in clinical trials, but with many limitations (different antibodies, tumour heterogeneity, specimens, and different thresholds of positivity). Other biomarkers, including tumour mutational burden and molecular signatures, are also developed and discussed in this review.

Deep learning (DL), often called artificial intelligence (AI), has been increasingly used in Pathology thanks to the use of scanners to digitize slides which allow us to visualize them on monitors and process them with AI algorithms. Many articles have focused on DL applied to prostate cancer (PCa). This systematic review explains the DL applications and their performances for PCa in digital pathology. Article research was performed using PubMed and Embase to collect relevant articles. A Risk of Bias (RoB) was assessed with an adaptation of the QUADAS-2 tool. Out of the 77 included studies, eight focused on pre-processing tasks such as quality assessment or staining normalization. Most articles (n = 53) focused on diagnosis tasks like cancer detection or Gleason grading. Fifteen articles focused on prediction tasks, such as recurrence prediction or genomic correlations. Best performances were reached for cancer detection with an Area Under the Curve (AUC) up to 0.99 with algorithms already available for routine diagnosis. A few biases outlined by the RoB analysis are often found in these articles, such as the lack of external validation. This review was registered on PROSPERO under CRD42023418661.