Radiomic Machine Learning Classifiers in Spine Bone Tumors: A Multi-Software, Multi-Scanner Study

• Published in: European journal of radiology Vol. 137; p. 109586
• Main Authors: Chianca, Vito, Cuocolo, Renato, Gitto, Salvatore, Albano, Domenico, Merli, Ilaria, Badalyan, Julietta, Cortese, Maria Cristina, Messina, Carmelo, Luzzati, Alessandro, Parafioriti, Antonina, Galbusera, Fabio, Brunetti, Arturo, Sconfienza, Luca Maria
• Format: Journal Article
• Language: English
• Published: Ireland Elsevier B.V 01.04.2021
• Subjects: Adolescent; Adult; Aged; Aged, 80 and over; Artificial Intelligence; Bone Neoplasms - diagnostic imaging; Child; Female; Humans; Machine Learning; Magnetic Resonance Imaging; Male; Middle Aged; Neoplasms; Radiomics; Retrospective Studies; Software; Spine; Young Adult; NEX; 0R; GLRLM; PBT; MRI; Spine; Artificial Intelligence; NGTDM; D-ANN; GLSZM; ROI; TA; GLCM; hCAD; Neoplasms; GLDM; Radiomics; TE; Magnetic Resonance Imaging; IOLB; TR; ML
• ISSN: 0720-048X; 1872-7727; 1872-7727
• DOI: 10.1016/j.ejrad.2021.109586

•Radiomics-based machine learning (ML) is promising for spine lesion classification.•ML was up to 94% accurate in distinguishing benign from malignant entities.•The best ML models were comparable to an experienced musculoskeletal radiologist.•Image pre-processing allowed for more stable performance across test datasets. Spinal lesion differential diagnosis remains challenging even in MRI. Radiomics and machine learning (ML) have proven useful even in absence of a standardized data mining pipeline. We aimed to assess ML diagnostic performance in spinal lesion differential diagnosis, employing radiomic data extracted by different software. Patients undergoing MRI for a vertebral lesion were retrospectively analyzed (n = 146, 67 males, 79 females; mean age 63 ± 16 years, range 8-89 years) and constituted the train (n = 100) and internal test cohorts (n = 46). Part of the latter had additional prior exams which constituted a multi-scanner, external test cohort (n = 35). Lesions were labeled as benign or malignant (2-label classification), and benign, primary malignant or metastases (3-label classification) for classification analyses. Features extracted via 3D Slicer heterogeneityCAD module (hCAD) and PyRadiomics were independently used to compare different combinations of feature selection methods and ML classifiers (n = 19). In total, 90 and 1548 features were extracted by hCAD and PyRadiomics, respectively. The best feature selection method-ML algorithm combination was selected by 10 iterations of 10-fold cross-validation in the training data. For the 2-label classification ML obtained 94% accuracy in the internal test cohort, using hCAD data, and 86% in the external one. For the 3-label classification, PyRadiomics data allowed for 80% and 69% accuracy in the internal and external test sets, respectively. MRI radiomics combined with ML may be useful in spinal lesion assessment. More robust pre-processing led to better consistency despite scanner and protocol heterogeneity.