Deep Learning-Based Screening for POLE mutations on Histopathology Slides in Endometrial Cancer
van den Berg, N.; Schoenpflug, L.; Horeweg, N.; Volinsky-Fremond, S.; Barkey-Wolf, J.; Andani, S.; Lafarge, M. W.; Oertft, G.; Jobsen, J. J.; Razack, R.; Gerestein, K.; Jonges, T.; de Kroon, C. D.; Nout, R.; Tseng, D.; Kuijsters, N.; Powell, M. E.; Khaw, P.; Shepherd, L.; Leary, A.; de Boer, S. M.; Kommoss, S.; van den Heerik, A. S. V. M.; Haverkort, M. A. D.; Church, D.; de Bruyn, M.; Smit, V. T. H. B. M.; Steyerberg, E.; Creutzberg, C. L.; Koelzer, V. H.; Bosse, T.
Show abstract
POLE sequencing for somatic mutations (POLEmut) guides adjuvant therapy in endometrial cancer (EC), but cost and infrastructural considerations lead to limited uptake. Omission of POLE testing leads to unnecessary exposure to radiotherapy and/or chemotherapy. We developed POLARIX, a multiple instance deep learning model with attention pooling, which predicts POLE mutation status from routine hematoxylin and eosin whole-slide images (WSIs). Trained on 2,238 cases from eleven EC cohorts, POLARIX showed clinical-grade discrimination across three external cohorts (Pooled: AUC=0.95, 95% CI: 0.91-0.98; n=68/481 POLEmut/POLEwt). Attention maps highlight POLE morphologies. Clinical applicability is demonstrated using predefined thresholds based on three resource scenarios. The most sensitive threshold ("Low") yields a test reduction of 77% (73%-81%) (sensitivity: 93% (85%-99%), specificity: 89% (87%-92%)). POLARIX is an interpretable and cost-efficient approach to reduce POLE testing in women with endometrial cancer, broadening access to precision oncology.
Matching journals
The top 6 journals account for 50% of the predicted probability mass.