Automatic pancreas segmentation using ResNet-18 deep learning approach

The accurate pancreas segmentation process is essential in the early detection of pancreatic cancer. situated abdominal cavity human body. contains pancreas, liver, spleen, kidney, and adrenal glands. Sharp smooth from this a challenging tedious job medical image investigation. Top-down approaches like Novel Modified K-means Fuzzy clustering algorithm (NMKFCM), Scale Invariant Feature Transform (SIFT), Kernel Density Estimator (KDE) algorithms were applied for days. Recently, Bottom-up method has become popular analysis cancer diagnosis. LevelSet used to detect cavity. deep learning, bottom-up approach performance better than another. Deep Residual Network (ResNet-18) sharp CT scan images. 18 layers are architecture ResNet-18. automatic kidney accurately extracted proposed images dataset 82 patients. 699 150 with different angles training testing purposes, respectively. ResNet-18 attains dice similarity index value up 98.29±0.63, Jaccard Index 96.63±01.25, Bfscore 84.65±03.96. validation accuracy 97.01%, loss rate achieves 0.0010. class imbalance problem solved by weight data augmentation.