Table 2 Experimental evaluation indicators.
Indicators | Description |
|---|---|
\(mAcc_{os} = \frac{TP + TN}{{TP + TN + FP + FN}}\) | The average proportion of correctly labeled pixels to total pixels in the three types of MRI images |
\(mPre_{os} = \frac{TP}{{TP + FP}}\) | The average ratio of adjudicated positive samples to true positive samples in the three subdatasets |
\(mRe_{os} = \frac{TP}{{TP + FN}}\) | The mean proportion of samples with positive predictions that also have positive true values across the three types of MRI images |
\(mF1_{os} = 2*\frac{{Pre_{os} *Re_{os} }}{{Pre_{os} + Re_{os} }}\) | The F1-Score is the harmonic mean of precision and recall, ranging from a minimum of 0 to a maximum of 1. A higher F1-Score indicates better model robustness |
\(mIOU_{os} = \frac{{S_{pre} \cap S_{truth} }}{{S_{pre} \cup S_{truth} }}\) | This metric represents the mean intersection-over-union (IoU) ratio between the segmented tumor regions in the three sub-datasets and the tumor regions manually segmented by doctors |
\(mDSC_{os} = 2 * \frac{{\left| {S_{pre} \cap S_{truth} } \right|}}{{\left| {S_{pre} } \right| + \left| {S_{truth} } \right| }}\) | This metric measures the mean degree of similarity between the tumor regions segmented by the model and those manually segmented by doctors across the three sub-datasets |
