An end-to-end-trained vision-language model for native-language prostate pathology report generation
2026-08-24 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors created a new system to help doctors write reports about prostate cancer biopsies in different languages, like German, without relying on English-based tools. They used an automated method to break down old reports into smaller parts linked to specific biopsy images, so the model could learn without extra work from people. Their model can accurately detect cancer and grade it based on biopsy images, performing similarly to existing approved tools. This approach allows hospitals to build language-specific reporting tools using their own past data.
prostate cancerbiopsystructured reportingvision-language modeltokenizerF1 scoreGleason gradingautomated data annotationlarge language modellatent-space augmentation
Authors
Christian Grashei, Fabian Gülhan, Maximilian Legnar, Fabian Stögbauer, Cleo-Aron Weis, Carolin Mogler, Peter Schüffler
Abstract
Prostate cancer is among the most frequently diagnosed malignancies worldwide, and structured reporting of each biopsy core burdens pathologists. Existing tools frame this as classification, leaving pathologists to assemble coherent reports, while many slide-level vision-language models rely on English-centric encoders that transfer poorly to other clinical languages. We present a slide-level framework generating prostate biopsy reports that is language-independent by construction: tokenizer and model are trained from scratch, demonstrated here in German. To address paired-data scarcity, an automated pipeline uses a locally deployed large language model to split composite reports into core-specific image-text pairs, yielding 17,344 pairs from 2,402 historical cases without manual annotation. Evaluated for clinical attributes rather than linguistic similarity, the model achieves 96.2% F1 for malignancy detection and 65.2% for Gleason grading, competitive with an FDA-cleared classifier. Grading is further validated on three external cohorts with latent-space augmentation. Institutions can thus train native-language reporting models on their own archives.