Fig. 1.

Overview of the multiple neural network framework for integrated segmentation and quality control. For simplicity, this illustration shows an example of 3 single independent neural networks. (A) A T1 map is analyzed by (B) independent segmentation models to output (C) single-model segmentations. Then, the single-model segmentations are passed to a label voting scheme to generate (D) combined-model segmentations. (E) In addition, the agreement of the segmentation can be visualized. (F) A DSC matrix is generated from both single model and combined-model segmentations for (G) DSC predictions with regression models. (H) The final segmentation is chosen based on the DSC prediction, and the corresponding predicted DSC is output as (I) the final quality control score.