Deep-Learning for High Quality and High Quantitative Ultrasonic Echo Imaging

This paper performs in simulations deep learning (DL) for high quality and high quantitative ultrasonic (US) echo imaging: (i) reduction of multiple echoes (multiple reverberations) and (ii) grading lobe echoes, (iii) separation of multiply crossed waves in US echo images, (iv) US attenuation correction imaging and (v) superresolutioned reflection and scattering imaging. In addition, (vi) segmentations of benign and malignant (cancerous) tumors in breast tissues are also performed. Clinical Relevance- This study about DL suggests the possibility of DL US segmentation for the automatic differential diagnosis about the human in vivo breast tumors in conjunction with the surrounding DL models.