Prostate cancer is one of the most frequently diagnosed cancers in men worldwide.1 Despite advances in imaging and biomarker detection, definitive diagnosis and risk assessment of prostate cancer still require histopathological examination of a patient’s tissue to accurately characterize the tumor, reveal its aggressiveness, and guide early intervention and proper treatment planning.
In a typical prostate cancer diagnostic workflow, pathologists determine the presence of cancer, assign a Gleason Grade, and evaluate whether tumors have adverse findings that can affect the diagnosis, such as signs of invasiveness (i.e., perineural invasion (PNI)) and poor prognosis features (i.e., Gleason 4 cribriform pattern). However, these workflows are time-consuming and subject to variability between pathologist assessments.
AI can streamline prostate cancer pathology workflows, accelerating the detection of high-risk features that predict metastatic potential and helping clinicians guide patients to the most appropriate treatments sooner. Below, we outline how AI decision-support models augment these workflows.
When prostate cancer is suspected (often due to elevated Prostate-Specific Antigen (PSA) levels or abnormal MRI findings), tissue samples are obtained via needle biopsy for histopathological examination.
Prostate sampling typically combines systematic and targeted biopsies, but the former approach is most commonly used for making an initial diagnosis. Still, systematic biopsies can miss cancer foci or under-sample the most aggressive areas, leading to grade misclassification, and posing a high risk of both over- and under-diagnosed prostate cancer.
On the other hand, newer, MRI-targeted core needle biopsies sample regions that appear suspicious based on imaging, resulting in a higher detection rate of high-grade cancers. Ultimately, combining systematic and targeted biopsies allows for increased tumor sampling and better detection of all grades of prostate cancer.
Contemporary prostate-biopsy protocols typically procure 10–18 core specimens. Each core must be examined in detail and assigned a Gleason grade; omission of even a single high-grade glandular pattern can reclassify the patient into a higher-risk category.
This workload affects both efficiency and reproducibility. Inter-observer concordance for borderline Gleason pattern 3 versus 4 is only about 60%, and subtle findings, such as perineural invasion or cribriform pattern 4, may be overlooked when time is limited.
AI-driven whole-slide analysis changes the game: it scans every pixel, flags perineural invasion and cribriform pattern 4, and delivers an annotated slide—transforming a tedious search into a focused confirmation.
By integrating PNI and G4-cribriform detection models into routine digital pathology, Aiforia’s Prostate Cancer Adverse Findings module automatically pre-screens each whole-slide image for high-risk patterns.
The models are part of the Aiforia® Prostate Cancer Suite, alongside the Gleason Grade Groups model, and align with the standard diagnostic sequence: (1) confirm tumor presence, (2) assign Gleason grade, and (3) flag adverse prognostic features. In doing so, it adds a targeted, pixel-level check for perineural invasion and cribriform pattern 4 without disrupting the established workflow.
When a digital biopsy slide is opened, the Suite simultaneously applies the Gleason Grade Groups and Adverse Findings AI models, producing a single composite overlay that displays tumor extent, grade group, pattern-4 percentage, perineural invasion, and cribriform pattern 4. One pass, one uniform assessment.
Aiforia® Prostate Cancer Suite’s integrated output accelerates sign-out: internal benchmarks show up to a 34% reduction in slide review time while maintaining full slide coverage and minimizing the risk of missed micro-lesions.
Each module can also run independently. Laboratories may deploy the PNI or cribriform detector alone for quality-control audits or retrospective studies, or add the full integrated run for routine diagnostics to mirror the step-by-step workflow pathologists already follow.
The Prostate Cancer Adverse Findings module is validated for both systematic and MRI-targeted biopsies, so it supports pathologists with whatever sampling strategy is used. Running in tandem with the Gleason Grade Groups model, it pre-screens every pixel in minutes, flags even minute tumor foci, and presents a single composite overlay. Internal benchmarks show up to a 34% cut in slide-review time while maintaining a cancer recall of 96% and above, keeping pace as biopsy volumes and complexity rise.
Each model returns cell-level masks and contours directly on the original slide. Pathologists can inspect, edit, or exclude any outlined gland, nerve, or cribriform nest, retaining full diagnostic control while meeting regulatory demands for algorithm explainability. The engine’s training on thousands of expert-annotated glands ensures consistent segmentation, with grades and percentages clearly traceable to the underlying pixels.
Tumor length per core, percentage pattern 4/5, cribriform load, and perineural-invasion status are extracted automatically and exported to structured LIS fields—eliminating manual tallies and transcription errors. Aiforia’s open web-based API slots into existing laboratory systems, so reports populate instantly and can even trigger downstream tests, delivering objective data for prognostic models without extra clicks.
Result visualization in Aiforia® Clinical Suite Viewer
Biopsy perineural invasion (PNI) nearly doubles the risk of lethal prostate cancer after adjusting for Gleason grade and tumor volume, and is consistently associated with extracapsular extension, biochemical recurrence, and cancer-specific mortality. Therefore, early, standardized detection is critical.
Aiforia® Prostate Cancer PNI identifies and annotates the affected nerve together with tumor infiltration at 92.6% accuracy, ensuring that this high-risk feature is not overlooked at the biopsy stage.
Likewise, cribriform pattern 4 signals aggressive disease and a higher likelihood of metastasis. After the tumor has been graded, Aiforia® Prostate Cancer G4 Cribriform highlights cribriform foci within Grade 4 cancers with 91.3 % accuracy, enabling pathologists to prioritize these cases for rapid clinical action.
Together, these adverse-finding detectors give pathologists pixel-level confirmation of the two key prognostic markers most closely linked to poor outcomes, without adding a single extra step to the routine diagnostic workflow.
Aiforia® Prostate Cancer Suite isn’t just a single-use tool. When integrated into diagnostic prostate cancer pathology workflows, it's an evolving, long-term solution. Upcoming modules like the extraprostatic extension and surgical-margin analysis can expand the scope of detectable prostate cancer features, further augmenting pathologists’ workflows and improving decision-making around adjuvant therapy or surveillance.
Learn more about Aiforia Clinical Suites →
Aiforia® Prostate Cancer Gleason Grade Groups, PNI, and G4 Cribriform models are CE-IVD marked for diagnostic use in EU and EEA countries, and for Research Use Only (RUO) and Performance Studies Only (PSO) in all other market areas.
References
1. Niu, Y. et al. (2022). The Role of Perineural Invasion in Prostate Cancer and Its Prognostic Significance. Cancers, 14(17), 4065. https://doi.org/10.3390/cancers14174065
2. Zareba, P. et al. (2017). Perineural Invasion and Risk of Lethal Prostate Cancer. Cancer epidemiology, biomarkers & prevention, 26(5), 719–726. https://doi.org/10.1158/1055-9965.EPI-16-0237