MRI-Based Radiomic to Assess Lipomatous Soft Tissue Tumors Malignancy: A Pilot Study - Archive ouverte HAL Accéder directement au contenu
Poster De Conférence Année : 2018

MRI-Based Radiomic to Assess Lipomatous Soft Tissue Tumors Malignancy: A Pilot Study

Résumé

Introduction: Among lipomatous soft tissue tumor, the noninvasive diagnosis between benign lipomas, borderline well differentiated liposarcomas (WDL)/atypical lipomatous tumors (ALT) and high grade subtype of liposarcomas is crucial since it directly drive the therapy strategy. Our aim is to develop a MRI-based radiomic method to classify between these three forms of tumors. Methods: 105 subjects with lipomatous soft tissue tumors with histology and fat-suppressed T1w contrast enhanced MR images available were retrospectively enrolled to constitute the database. According to histology, 3 groups have been constituted: benign (including deep lipomas, n = 23); intermediate (including ALT and WDL, n = 41); and malignant (including high grade liposarcomas: myxoid, dedifferentiated, and pleomorphic, n = 41). MR images were obtained from 56 different centers with non-uniform protocols (3 fields: 1.0T, 1.5T, and 3.0T, with 18 different MR systems commercialized by 4 vendors). Images were automatically loaded on an in-house software and tumor was manually segmented by two observers blinded to histology to extract the radiome. The radiome included 87 features (size, shape, intensity distribution, image domain and frequency domain textures features). Radiomes were next mined with a supervised machine learning approach to build a decisional algorithm based on two radiomic models. Results: To classify between benign and (intermediate + malignant) groups, the 12th order model gave the best performance (AUROC: 0.959; sensitivity: 89%; specificity: 95.7). To classify between intermediate and malignant groups the 17th order model gave the best diagnosis performances (AUROC: 0.907; sensitivity: 85.4%; specificity: 90.2%). Conclusion: These results show that the evaluation of lipomatous tumor malignancy is feasible using a routinely used MRI acquisition in clinical practice. These encouraging results need to be further confirmed on another prospective or existing application cohort.

Domaines

Imagerie Cancer
Fichier non déposé

Dates et versions

hal-01990614 , version 1 (23-01-2019)

Identifiants

  • HAL Id : hal-01990614 , version 1

Citer

Benjamin Leporq, Amine Bouhamama, F. Lame, Catherine Bihane, Michaël Sdika, et al.. MRI-Based Radiomic to Assess Lipomatous Soft Tissue Tumors Malignancy: A Pilot Study. Forum de la recherche en Cancérologie 2018, Apr 2018, Villeurbanne, France. ⟨hal-01990614⟩
44 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More