A nomogram based on magnetic resonance imaging features for preoperative risk stratification in patients with endometrial cancer

Objective: This study aimed to develop a nomogram; based on age, tumor MRI morphological measurements and ADC Histogram analysis measurements to preoperatively distinguish low – non-low risk and high – non-high risk endometrioid endometrial cancer patients according to ESMO risk assessment criteria. Materials and Methods: Our study is a retrospective, descriptive case-control study. Endometrial cancer patients who underwent surgery within the first 30 days after imaging at Afyonkarahisar University Faculty of Medicine, Health Application and Research Center, between 01.01.2017 and 01.09.2023, were included. Out of 252 total patients, it was determined that 82 had complete imaging, surgery, and pathology results. Only endometrioid type endometrial cancers were included. Age, findings obtained in pathology results, tumor orthogonal diameters from lower abdominal MRI examinations including T2, Fat-saturated T1 contrast, and Diffusion/ADC series, tumor and uterus sagittal area measurements, and tumor/uterus area ratios were noted. Finally, volumetric tumor segmentation was performed from ADC maps, and histogram analysis resulted in minADC, maxADC, meanADC, medianADC, percentile10, percentile90, entropy, kurtosis, skewness values. Patients were classified according to ESMO criteria. Significant parameters between groups were determined using basic statistical tests, and the best model was created using binary logistic regression tests considering multicollinearity. A nomogram was obtained for visualization of the model. PASW Statistic 18, Jamovi, and R Software were used for statistical analysis, Slicer software for volumetric segmentation and histogram analysis, and p<0.05 was considered statistically significant. Results: Logistic regression models for the low – non-low risk group included parameters APSag, TAOSag, Median, Skewness; for the high – non-high risk group, APSag, TAOSag, Skewness were determined. Correct classification rates were 86.6% and 70.7%. The AUC, sensitivity, and specificity values obtained from ROC curves were 0.922, 0.934, 0.667 for the low – non-low risk group, and 0.766, 0.827, 0.500 for the high – non-high risk group. The nomogram visual was appended. Conclusion: Comparative to literature, the nomogram integrating ADC Histogram analysis with MRI morphological measurements exhibited a considerable level of stratification in discriminating low – non-low risk groups for endometrioid endometrium cancer. Keywords: Endometrial cancer, risk stratification, nomogram, ADC histogram analysis, MRI.