A nomogram based on radiomics intermuscular adipose analysis to indicate arteriosclerosis in patients with newly diagnosed type 2 diabetes

Figure 1. Inclusion and exclusion criteria of our study.

Figure 2. Flowchart of the development and evaluation of the clinical model, radiomics model, and clinical-radiomics combined model. LR, linear regression; ROI, regions of interest; GLCM, co-occurrence matrix; GLSZM, grey-level size zone matrix; GLRLM, grey-level run length matrix; GLDM, grey-level distance-zone matrix; NGTDM, neighbourhood grey-tone difference matrix.

Figure 3. Model construction and evaluation. Nomograms of the combined model to indicate plaque presence (A) and plaque severity (B). Calibration curves of the nomogram indicating plaque presence for the training set (C) and the validation set (D). Calibration curves of the nomogram indicating plaque severity for the training set (E) and the validation set (F). The x-axes represent the probability of plaque presence or severity evaluated by the combined models, and the y-axes represent the actual rate of plaque presence or severity. The diagonal dotted lines represent perfect predictions by ideal models, while the solid lines represent the discrimination abilities of the nomograms, of which closer fits to the diagonal dotted lines represent better evaluations.

Figure 4. Confusion matrixes of the three models. Model 1, clinical model; Model 2, radiomics model; Model 3, clinical-radiomics combined model. The x-axes represent the predicted labels, and the y-axes represent the actual labels. Four quadrants clockwise from the upper left hand refer to true negatives, false positives, false negatives, and true positives, respectively. The false-positive and false-negative rates in model 3 were lower than model 1 in both the training sets and validation sets.

Figure 5. The receiver operating characteristic (ROC) curves of the three models. Model 1, clinical model; Model 2, radiomics model; Model 3, clinical-radiomics combined model. In all subjects included in the study, model 3 demonstrated better indicative efficacy than model 1 and similar to model 2, with an AUC of 0.934 in the training set (A) and an AUC of 0.933 in the validation set (B). In subjects with high plaque burden, model 3 demonstrated the best indicative efficacy than the other two models, with an AUC of 0.824 in the training set (C) and an AUC of 0.717 in the validation set (D).

Figure 6. Decision curves analysis for the three models. Model 1, clinical model; Model 2, radiomics model; Model 3, clinical-radiomics combined model. The x-axes represent the threshold probability, and the y- axes represent the net benefit. For all subjects included in the study, model 3 and model 2 had better performance than model 1 in both the training set (A) and the validation set (B). However, for subjects with high plaque burden, model 3 had higher efficacy than the other two models in the training set (C) and no significant differences in the validation set (D).