Discriminative and predictive power of radiomics features from preoperative MRI studies in patients with supraspinatus lesions.
Authors
Affiliations (7)
Affiliations (7)
- Fondazione Policlinico Universitario Campus Bio-Medico, Via Alvaro del Portillo, 200, Roma, 00128, Italy. [email protected].
- Resarch Unit of Orthopaedic and Trauma Surgery, Department of Medicine and Surgery, Università Campus Bio-Medico di Roma, Via Alvaro del Portillo, 21, Roma, 00128, Italy. [email protected].
- Fondazione Policlinico Universitario Campus Bio-Medico, Via Alvaro del Portillo, 200, Roma, 00128, Italy.
- Intelligent Health Technology for Health and Wellbeing lab, Department of Engineering, Università Campus Bio-Medico di Roma, Via Alvaro del Portillo, Rome, 21 - 00128, Italy.
- Resarch Unit of Orthopaedic and Trauma Surgery, Department of Medicine and Surgery, Università Campus Bio-Medico di Roma, Via Alvaro del Portillo, 21, Roma, 00128, Italy.
- Musculoskeletal Translational Innovation Initiative, Carl J. Shapiro Department of Orthopaedic Surgery, Beth Israel Deaconess Medical Center, Harvard Medical School, 330 Brookline Avenue, RN123, Boston, MA, 02215, USA.
- Unit of Measurements and Biomedical Instrumentation, Department of Engineering, Università Campus Bio-Medico di Roma, Via Alvaro del Portillo 21, Roma, 00128, Italy.
Abstract
Relying solely on MRI scans may prove insufficient for obtaining a complete view of a patient's supraspinatus health status comprising both objective and subjective disability level. On the other hand, while Patient Reported Outcome Measures (PROM) capture patients' personal experiences, they may not provide objective data.This study assessed the correlation level between preoperative MRI radiomics features and pre-and post-operative PROM scores in patients diagnosed with supraspinatus lesions. 44 patients who underwent arthroscopic repair for full-thickness Rotator Cuff Tears (RCT) of any grade between January 2019 and January 2022 in a single surgical centre were enrolled in the study. All coronal T2-weighted MRI scans were evaluated independently by two orthopedic surgeons blinded to the patients' clinical information, and the radiomics features were extracted. The strength of linear and non-linear relationships were assessed respectively through Fisher's statistical test applied to linear regression models, namely F-statistics, and through a model-free AI-based Mutual Information estimator. Together, these methods provided a comprehensive feature selection framework, with overlapping results reinforcing feature relevance and differences revealing complementary patterns in the radiomics-PROM relationship. Besides, they allowed us to construct an integrated ranking of radiomic features based on their strength of association with clinical outcomes, from which we selected those ranking in the top 10 most important of at least three PROMs.The radiomics features most frequently related to the patient's health condition were Maxi- mum2DDiameterColumn, Flatness, LeastAxisLength, and features gathered from the Gray Level Co-occurrences Matrix (GLCM), like Correlation, Maximum Correlation Coefficient (MCC), and Informational Measure of Correlation 2 (Imc2).Identifying crucial visual features pertaining to the patient's health status highlighted the potential for radiomics-guided personalized medicine in rotator cuff surgery. Incorporating both linear and non-linear analyses adds depth to the findings, establishing a strong groundwork for future advancements in the field.