Background and aims: The development of
artificial intelligence (AI) is essential for deployingcommunity-wide,
prostate cancer diagnosis. Prediction tools widely used and well-validateddepend on standard, readily available clinical and pathological parameters, but do not includebiomarkers, which can provide valuable insights into prediction or treatment options. Combiningtraditional prediction methods with systems pathology may be provided a more personalized riskassessment of clinically relevant outcomes of prostate cancer. Across a range of disease prevalence,AI systems can deliver the main benefits of biopsy avoidance while maintaining highspecificities.In this review, we examined current developments in precancerous lesion detection and diagnosisfor prostate cancer.Methods: A review was carried out by two reviewers independently and manually searching Englishdatabases (PubMed, Scopus, and Web of Science) for data till March ۲۰۲۳. After the qualityscreening, ۳۴ articles was made for further analysis. Database were searched using the terms ‘artificialintelligence’, ‘prostate cancer’, ‘precision diagnosis’, and ‘precision diagnosis’. Inclusioncriteria for paper selection were: ۱) Paper must be peer reviewed. ۲) Journals on which paperspublished must be either PubMed, Scopus, or Web of Science indexed. ۳) The paper should useonly AI techniques. Exclusion criteria for paper selection were: ۱) Duplicate studies in differentdatabases. ۲) Study which is less cited by other peer reviewed papers. ۳) MSc and PhD papers.Results: In the literature, a wide number of machine learning techniques have been applied tobiopsy material, including linear models, support vector machines, decision trees, and deep learningmodels for
prostate cancer diagnosis. In the following comprehensive review article, we focuson diagnostic (PHI®, ۴K score, SelectMDx®, ConfirmMDx®, PCA۳, MiPS, ExoDX®, mpMRI)biomarkers that are in widespread clinical use and are supported by evidence. In addition,we discussed new biomarker-driven diagnosis for advanced
prostate cancer that have been obtainedusing
artificial intelligence such as TELO۲, ZMYND۱۹, miR-۱۴۳, miR-۳۷۸a, cg۰۰۶۸۷۳۸۳(MED۴), and cg۰۲۳۱۸۸۶۶ (JMJD۶; METTL۲۳).Conclusion: Such a variety of cancer molecular and clinical data calls for advancing the interoperabilityamong AI approaches, with particular emphasis on the synergy between discriminativeand generative models that we discuss in this work with several examples of techniques and applications.To improve the predictive power of potential diagnostic biomarkers, experiments mustbe carefully designed.