“
TL;DR
A deep learning model called NaroNet, trained on multiplexed immunofluorescence images of tumor tissue, predicted recurrence in low-grade, early-stage endometrial cancer with high accuracy, outperforming molecular subtyping.
NPJ Digital Medicine · 12 authors, 11 centres
This summary was generated by AI from a single paper. It has not been reviewed by a clinician and is not clinical advice. Verify against the source before acting on it.
A deep learning model called NaroNet, trained on multiplexed immunofluorescence images of tumor tissue, predicted recurrence in low-grade, early-stage endometrial cancer with high accuracy, outperforming molecular subtyping.
This retrospective cohort study developed a weakly-supervised, multilevel deep learning model (NaroNet) to predict recurrence in low-grade, early-stage endometrial cancer using multiplexed immunofluorescence images of 489 tumor cores from 250 patients. The model was trained without manual annotation, learning from patient recurrence labels to identify tumor-immune interrelations at local, neighborhood, and tissue area levels. Tissue areas identified as 'inflamed' (hot tumors) were associated with no recurrence, while 'noninflamed' (cold tumors) areas were linked to recurrence. Limitations include reliance on only two tumor cores per patient, which may not fully capture spatial heterogeneity, and potential batch-to-batch variations in multiplexed IF processing. The findings suggest this method could accurately assess recurrence risk, offering a tool for clinical decision-making.