Skip to main content
. 2020 Dec 31;17(9):6489–6498. doi: 10.1109/TII.2020.3048391

Fig. 1.

Fig. 1.

Graphical overview of the optimization scheme of CovTANet: Tri-level attention-based segmentation network (TA-SegNet) extracts the slicewise lesion segmentation mask and representational features of the corresponding CT-volume which are employed later for the joint optimization of severity prediction and diagnosis. Separate tri-level attention units (TAUs) are employed to enhance the diagnostic features and severity-based features in the joint optimization process.