Evaluation of radiomics in improving the diagnostic efficacy of whole body 99m Tc-MDP bone scintigraphy

Abstract

Abstract [Objective] to establish and verify the prediction model by using radiomics, and explore whether radiomics can improve the diagnostic efficiency of whole-body 99mTc MDP bone scintigraphy. [Methods] 79 patients who completed whole body bone scintigraphy were retrospectively analyzed. The 19 regions of interest were merged into one region of interest on the anterior posterior and posterior anterior positions of whole body bone scintigraphy images, and the consistency of the delineation results was evaluated. The 99mTc-MDP bone scintillation image features were extracted by Pyradiology 1.23.1. They were randomly divided into training group and test group according to 7:3. Univariate logic analysis and variable of stepwise selection method were used to screen the characteristics. Use forest to build a model from the selected features and test the model characteristics. The R language was selected to draw the working characteristic curve of the subjects to determine the performance of the machine learning model. The accuracy, sensitivity, specificity and area under the curve were calculated, which were statistically significant (P < 0.05). [Results] the consistency of anterior posterior and posterior anterior ROI of 99mTc-MDP whole body scintigraphy was 0.98. There were significant differences in the regional characteristics of ROI between the bone metastasis group and the non bone metastasis group. Use variable, threshold 1, to further reduce the number of features to 8. The sensitivity, specificity and accuracy of the training group and the test group were 88.9%, 96.4% and 92.7%, 91.7%, 91.7% and 91.7%, respectively. [Conclusion] the accuracy of the prediction model of bone metastases established by radiomics method is better than the deep learning method, which can improve the whole-body diagnostic efficiency of 99mTc-MDP bone scintigraphy image, especially for small sample data, improve the diagnostic accuracy and reduce the clinical workload, which has good application and promotion value and potential.