A-study-of-attention-information-from-transformer-layers-in-hybrid-medical-image-segmentation-networks