A shield automated model for medical image segmentation using successive image Categorical Network Model

Medical Image Classification and segmentation are the crucial procedures in clinical diagnosis. Typically, regions of interest (ROI) are properly split to extract significant data for additional disease categorization. These techniques are time-consuming and computationally complex. This study proposed a novel Successive Image Categorical Network Model (SICNet) that correctly recognizes image and offers high-quality segmentation for every medical object. While the classification task employed a ResNet101 bottleneck module for feature fusion, the segmentation challenge used a reverse addition module to combine areas in boundary cues and global maps in high-resolution features. For imaging modalities analysis, three online accessible datasets we evaluated SICNet’s performance against existing architectures. In the diagnosis of tumors, our suggested model performed better than all 25 radiologists and prevailing models on all three datasets. The anticipated SICNet model gives 96%, 95% and 96.5% accuracy for the provided datasets. Similarly, the datasets yield an AUC of 91%, 92.7% and 93%, respectively. The time needed for executing these three datasets are 0.25s, 0.27s and 0.29 s, respectively.