Pilot Clinical and Radiomic Analysis of Deep Infiltrating Endometriosis of the Parametrium Using Shannon Entropy: A Retrospective Cohort Study

OBJECTIVE: Ultrasound techniques for diagnosing deep infiltrating endometriosis (DIE) currently lack a quantitative method to assess microstructural heterogeneity in relation to diagnosis and clinical symptoms. This study evaluates Shannon entropy-based radiomics for differentiating DIE lesions from adjacent tissue and correlating these features with pain severity. METHODS: In this retrospective cohort study (2020-2024), fertile women with histologically confirmed parametrial endometriosis and high-quality ultrasound images were evaluated. Pain was measured using a 10-cm visual analog scale. Two 25 mm² regions of interest (ROIs) were defined: one within the DIE nodule and one in the adjacent perilesional tissue. Each ROI was analyzed with the Shannon entropy algorithm to assess tissue heterogeneity. RESULTS: In this study, 148/663 women (22.3%) were diagnosed with parametrial DIE, of whom 52 underwent surgery and were included in the study. DIE was localized in 92.3% in the posterior, in 5.8% in both posterior and anterior, and in 2% in the anterior parametrium. Primary symptoms were menstrual pain (88%) and pain at intercourse (63%). ROI entropy of DIE lesion was lower than that of the perilesional ROI (p = 0.05). The area-under-the-curve (AUC) of Shannon entropy for endometriotic vs. adjacent tissue was 91.36% (95% CI: 84.5%-98.21%). The ROI entropy of the DIE lesion exhibited an inverse correlation with menstrual pain (rho -0.46, CI 95% -0.74/-0.12, p = 0.025), and that of perilesional tissue with chronic pelvic pain (rho -0.41, 95% CI -0.73/- 0.04; p = 0.072). CONCLUSION: Entropy analysis distinguishes DIE lesion from perilesional tissue and significantly correlates with menstrual pain.