{"paper_id":"bed32de4-66bf-47f5-be43-3445c25f34ec","body_text":"ECR 2025 / C-16010\nCan AI Help Radiologists Discriminate Between Active and Fibrotic Lesions in Deep Endometriosis Through MR Imaging? A Radiomic Features Study\nCongress:\nECR 2025\nPoster Number:\nC-16010\nType:\nScientific Exhibit\nKeywords:\nArtificial Intelligence, Genital / Reproductive system female, Pelvis, MR, Segmentation, Tissue characterisation\nAuthors:\nV. Lucidi, M. C. Di Giovanni, N. Curti, S. Peluso, D. Raimondo, M. Arcilesi, G. Castellani, R. Seracchioli, C. Mosconi\nDOI:\n10.26044/ecr2025/C-16010\nPurpose\nDeep endometriosis is a chronic inflammatory diseases associated with pelvic organ involvement (usually uterus and sigma-rectum), severe pain and infertility [1]. Its prevalence is estimated to be 1% -2%.Deep endometriosis lesions are generally distinguished in active and fibrotic tissue, and such distinction withholds clinical importance as active lesions might present a good response to hormonal treatment, whereas fibrotic lesions are stable and generally do not respond to hormonal therapy.MRI features of deep endometriosis lesions include a hypointense signal in T2-weighted imaging with or without hyperintense...\nMethods and materials\nIn this monocentric prospective study, 63 patients with deep endometriosis who underwent MRI examination and received consecutive surgical treatment were enrolled. During surgery, the distribution of all the endometriosis lesions was noted and successively correlated with their histological specimen (active vs fibrotic).The MRI protocol (1,5 Tesla) included T2-weighted acquisitions in each plane (axial, coronal and sagittal) without fat suppression, T1-weighted axial acquisitions with andwithout fat suppression, post-contrast enhancement T1-weighted acquisitions with fat suppression and Diffusion-weighted axial acquisitions after a preparation of rectal distension with aquagel...\nResults\nEven though all the segmented lesions presented hypointense signal in T2-weighted imaging, which according to literature is a feature that correlates with fibrotic nature, as many as 77.1% (145) corresponded to active tissue, and only 22.9% (43) of those lesions have been histologically confirmed as fibrotic. Additionally, both phenotypes of lesions could present hematic signal in T1-weighted imaging with fat suppression without a clear correlation with histological activity.Regarding the radiomic analysis, thirteen features were extracted, belonging to the category of Haralick features, calculated from the...\nConclusion\nIn MR imaging, the radiologist should describe deep endometriosis lesions by listing their radiological features and morphology without suggesting a correlation with their histological nature, therefore without using the terms “fibrotic” nor “active”, as it was demonstrated that hematic signal in T1w wasn't exclusively correlated with histological activity of glandular tissue, nor hypointense signal was exclusively correlated with fibrotic tissue.Nonetheless, our radiomic model might help to establish a machine learning algorithm to predict to which group an endometriotic deep lesion might belong without or before...\nPersonal information and conflict of interest\nV. Lucidi:\nNothing to disclose\nM. C. Di Giovanni:\nNothing to disclose\nN. Curti:\nNothing to disclose\nS. Peluso:\nNothing to disclose\nD. Raimondo:\nNothing to disclose\nM. Arcilesi:\nNothing to disclose\nG. Castellani:\nNothing to disclose\nR. Seracchioli:\nNothing to disclose\nC. Mosconi:\nNothing to disclose\nReferences\nWang PH, Yang ST, Chang WH, Liu CH, Lee FK, Lee WL.Endometriosis: Part I. Basic concept. Taiwan J Obstet Gynecol. 2022 Nov;61(6):927-934. doi: 10.1016/j.tjog.2022.08.002. PMID: 36427994.\nBazot M, Bharwani N, Huchon C, Kinkel K, Cunha TM, Guerra A, Manganaro L, Buñesch L, Kido A, Togashi K, Thomassin-Naggara I, Rockall AG. European society of urogenital radiology (ESUR) guidelines: MR imaging of pelvicendometriosis. Eur Radiol. 2017 Jul;27(7):2765-2775. doi: 10.1007/s00330-016-4673-z. Epub 2016 Dec 5. PMID: 27921160; PMCID: PMC5486785.\nKoninckx PR, Ussia A, Adamyan L, Wattiez A, Donnez J.Deep...","source_license":"CC0","license_restricted":false}