PathXAI preprint reports gains in evidence-linked pathology reporting

Researchers report in a preprint that PathXAI improved automated pathology reporting by combining tissue measurements with stained tissue images. On its public evaluation cohort, PathXAI-Report exceeded Qwen + H&E in cancer family match rate by 4.4 percentage points and in finding F1 by 5.7 percentage points. The system uses tissue segmentation to guide diagnostic reasoning, targeted reinspection and structured report writing. It generates evidence-linked reports with a self-hosted pretrained vision-language model without updating its weights. Evaluation covered 1,200 public and 151 private cases; the reported gains concern automated comparisons with reference diagnoses and documented morphology.