Friday, August 14, 2026

AI Guest Author: How Far Away Are AI-Generated Path Reports? Closer Than You Knew! (Ver2)

 Only seven days ago, we published a white paper asking, “How Far Away Are AI-Generated Pathology Reports?” Since then, we encountered several important papers, commercial programs, and industry reports that we had not incorporated in the original review—and they materially change the answer. In particular, current prostate-biopsy AI is considerably closer to assembling and prepopulating a pathology report than we had appreciated. We therefore went back to the evidence and substantially revised the white paper. The original August 7 version is available here; the new August 14 edition reflects this much more advanced pathology landscape.

Find the new paper: >>   here   <<



Summary

Medical imaging AI is moving beyond detection and measurement toward systems that assemble the diagnostic work product. The clearest regulatory example is DeepHealth SMART-B, an FDA-cleared breast ultrasound system that analyzes lesions and generates report findings and impressions for radiologist review. Pathology is approaching the same boundary faster than expected, particularly in prostate biopsy. Current commercial systems can identify cancer, assign Gleason patterns and Grade Groups, measure tumor length and percentage involvement, and transfer structured results directly into reporting interfaces. European products from Aiforia and Ibex already describe first-read and automated-reporting workflows, while U.S. FDA authorizations remain more conservative and largely adjunctive or second-read.

Research systems push further. HistoGPT generates dermatopathology reports from whole-slide images; TITAN and PRISM2 demonstrate report generation or completion of structured pathology fields. Importantly, the first useful pathology reports may not require an unconstrained language model. A validated image-analysis stack can produce structured diagnostic facts, and a deterministic template can convert those facts into an editable draft linked to supporting image evidence. The remaining barriers are less about fluent writing than reliable case assembly, uncommon findings, laboratory variability, workflow integration, safety, provenance, and FDA authorization. Prostate needle biopsy may be the first major U.S. test case.