Skip to main content
Fig. 5 | EJNMMI Research

Fig. 5

From: Dynamic planar scintigraphy for the rapid kinetic measurement of myocardial 123I-MIBG turnover can identify Lewy body disease

Fig. 5

Representative results of machine learning classifier models for iLoss and iUp/iLoss to discriminate patients with Lewy body disease (LBD) from those with non-LBD. After correction for class imbalance using SMOTE, 210 coordinates of iUp and iUp/iLoss, which represent 105 LBD patients (squares in red) and 105 oversampled non-LBD patients (squares and triangles in blue), are shown (A). An example portion allocated for testing (63 samples, 30%) is shown with the independent cut-off values of iLoss and iUp/iLoss (B). Decision boundaries (thick lines) and predicted probabilities (thin lines) were generated from single runs of the SVM classifier models with the linear kernel (C) and the radial basis function kernel (D)

Back to article page