par Larhmam, M.A.;Mahmoudi, Saïd;Drisis, Stylianos ;Benjelloun, Mohammed
Référence Lecture notes in computer science, 10814 LNBI, page (198-211)
Publication Publié, 2018
Article révisé par les pairs
Résumé : This paper presents a learning based approach for the classification of pathological vertebrae. The proposed method is applied to spine metastasis, a malignant tumor that develops inside bones and requires a rapid diagnosis for an effective treatment monitoring. We used multiple texture analysis techniques to extract useful features from two co-registered MR images sequences (T1, T2). These MRIs are part of a diagnostic protocol for vertebral metastases follow up. We adopted a slice by slice MRI analysis of 153 vertebra region of interest. Our method achieved a classification accuracy of 90.17% ± 5.49, using only a subset of 67 relevant selected features from the initial 142.