The Role of Histogram Analysis in the Radiological Evaluation of Endometrial Cancer

Abstract Purpose The success of histogram analysis in T2-W imaging in predicting endometrial cancer will be investigated. Methods The study analyzed pelvic MRIs of 32 patients who underwent surgery for endometrial cancer (January 2016–October 2024) and 38 controls with pelvic MRIs available for other indications. Texture parameters (area, mean, variance, skewness, kurtosis, 1st, 10th, 50th, and 99th percentiles, as well as minimum and maximum luminance) were assessed using the MaZda software (version 4.6). In the patient group, regions of interest (ROIs) were selected from sagittal T2-weighted images where the tumor was most prominent, while midsagittal images were used for controls. Data from histogram analyses of T2-W images were compared using the independent t-test and Mann–Whitney U test, with ROC and linear discriminant analyses (LDA) also performed. Results Results revealed an average age of 61.4 ± 9.9 years in the patient group and 59.3 ± 10.5 years in the control group. Among the patients, 20 cases were FIGO stage 1, 9 were stage 2, and 3 were stage 3. Statistically significant differences were observed for texture parameters such as area, mean, variance, 1st, 10th, 50th, and 99th percentiles, as well as minimum and maximum luminance (p = 0.003 for variance, p < 0.001 for others), while skewness and kurtosis showed no significant differences (p = 0.804 and p = 0.106, respectively). ROC analysis demonstrated a high AUC for area (0.907), with sensitivity and specificity of 0.844 and 0.789, respectively. LDA identified area, mean, and variance as key predictors, achieving classification accuracies of 87.1% and 82.9% for original and cross-validated datasets, respectively. Conclusion The significant differences found in texture parameters obtained from T2-W histogram analysis between cases with endometrial cancer and control group suggest that histogram analysis may be useful in improving diagnostic accuracy.