Fig. 5: Combined framework pipeline – Auto-Prognose-CNNattention. | Communications Medicine

Fig. 5: Combined framework pipeline – Auto-Prognose-CNNattention.

From: A convolutional attention model for predicting response to chemo-immunotherapy from ultrasound elastography in mouse tumor models

Fig. 5

The SWE and B-mode parts are first identified in the ultrasound images. B-mode is then used as input for the tumor segmentation model and when the tumor region is segmented, it is transferred to the elastography (SWE) images. The tumor area is finally cropped and used as input for the prognostic prediction model. In the first example of the figure, the model predicted the tumor as stable with the probability of 78% and in the second example the model predicted the tumor as responsive with the probability of 99%.

Back to article page