Publication: Radiomics features of magnetic resonance imaging for prediction of survival in osteosarcoma patients
Loading...
Date
2021
Authors
Alshantti, Khalilalrahman Ibrahim Hassan
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Introduction: Osteosarcoma is a primary bone tumour associated with a poor prognosis, and prediction of overall survival of osteosarcoma can help in personalizing the patient's management. Recently many studies have shown that the tumour texture, the heterogeneity of tumors, and the intensity may have the potential to predict the prognosis or the survival of patients with osteosarcoma. The aim of this study was to investigate the prognostic value of Magnetic resonance imaging (MRI) texture, shape and intensity parameters in patients with osteosarcoma to develop a radiomics nomogram by combining clinical factors and the radiomics features. The radiomics features were extracted from the pre-treatment magnetic resonance images of T1WI, T2WI and T1 post contrast. Upon many features were extracted, the most significant features were selected then the Kaplan-Meier survival analysis was performed. Methodology: This retrospective study was conducted in Hospital Universiti Sains Malaysia (HUSM), Kelantan, from January 2000 until December 2010 on 60 patients were diagnosed within that period. MRI images of T1WI, T2WI and T1 post Gad sequences of patients who fulfilled the inclusion criteria were retrieved from the PACS system. Tumour segmentation and features extraction using MATLAB software. The Univariate analysis was done for features selection. A final model was built by integrating the significant radiomic features with the significant clinical data. Data were split into training and test data set in 70:30 ratio, randomly selected. Results: Total of sixty subjects (n=60) with a mean age of 17.39 years old were included in this retrospective study. Overall survival probability among the subjects was 45.7%. Cox proportional hazard regression for the patient clinical data and the extracted radiomics features showed no statistically significant of most of the radiomics fractures with (P > 0.05) as well as most of the radiomics features except 3 radiomic features extracted from T2WI. Multiple logistic regressions were done with increasing the P value to < 0.25, as results of that features and clinical data with P value < 0.25 were included in the final model. The sensitivity, specificity and accuracy of the model were 70%, 77.8% and 73.7. The positive and negative predictive values were 77.8% and 70% respectively. Conclusions: The study showed the radiomic features were statistically significant compared to the clinical data which can play a major role in prediction of the patient's survival. However due to the limited data of one centre the results of the study were not significantly high. Multicentric study with large data should be performed to obtain significant results and be useful in the future clinical patient management.
Description
Keywords
Osteosarcoma , survival prediction