Livernet based segmentation of lesions from computed tomography scan for liver tumor detection

Millions of lives could be saved annually if liver tumours could be detected early with computed tomography. But it's a huge strain for radiologists to read hundreds or even tens of these CT scans. Therefore, developing an autonomous, rapid, and reliable method of reading, detecting, and assessing CT scans is important. However, extracting the liver region from CT scans is a bottleneck for any approach. This paper introduces a three-part automatic process. Initial processing includes noise suppression and image enhancement. Optimized Bi-lateral Filtering is used to carry it out; in this case, the process's control parameters are optimized using the Monarch butterfly optimization method. After that, automatic liver segmentation and lesion identification are performed. Mask-Region-based Convolutional Neural Network segment liver from the pre-processed images. Then a new generator network named LiverNet is used to detect tumors within the liver. Finally, an Enhanced Swin Transformer Network employing Adversarial Propagation distinguishes between malignant and benign liver lesions. Positive developments were discovered as a result of the inquiry. Expert results are associated with the consequences of segmentation and analysis. The classifier makes a relatively accurate tumour differentiation and gives the radiologist a second opinion.