An automatic segmentation of breast ultrasound images using U-Net model

Medical imaging, like ultrasound, gives a good visual picture of how an organ works. However, radiologist has hard time and takes long to process these images, which delays the diagnosis. Several automated methods for detecting segmenting breast lesions have been developed. Nevertheless, due ultrasonic artifacts intricacy lesion forms locations, segmentation or tumors from ultrasonography remains open issue. image seen breakthrough thanks deep learning. U-Net is most noteworthy network in this regard. The traditional design lacks precision when dealing with complex data sets, despite its exceptional performance multimedia medical images. To reduce texture detail redundancy avoid overfitting, we suggest developing architecture by including dropout layers after each max pooling layer. Batchnormalization binary cross-entropy loss function were used preserve tumor features edge attributes while decreasing computational costs. We ultrasound dataset 780 images normal, benign, malignant tumors. Our model showed superior results pictures compared previous neural networks. Quantitative measures, accuracy, IoU values utilized evaluate suggested model?s effectiveness. 99.34% 99.60% accuracy IoU. imply that augmented high diagnostic potential clinic since it can correctly segment lesions.