Exploiting deep transfer learning based precise classification and grading of renal cell carcinoma using histopathological images

Renal cancer is a key reason for cancer-related deaths among males worldwide. Earlier diagnosis of renal cancer is critical since it can considerably increase the chance of survivability. However evaluating the histopathological renal tissue is a tedious process and usually, this is manually done by the pathologist, resulting in a strong possibility of misdiagnosis or misdetection, particularly in the earlier phases, and susceptible to inter-pathologist variations. The advancement of automated histopathological diagnoses of renal cancer could significantly decrease the bias and offer correct classification of disease though the pathology and microscopy nature are more complicated and complex. Current researchers recommend that clinicians successfully implement the classification task by investigating the image texture feature of RCC from computed tomography (CT) data. However, image feature detection for RCC grading frequently depends on a physical process that is time-intensive and error-prone. Therefore, this article develops an Exploiting Deep Transfer Learning based Precise Classification and Grading of Renal Cell Carcinoma (EDTL-PCGRCC) method using Histopathological Imaging. The projected EDTL-PCGRCC methods inspect the histopathological images for the classification and detection of RCC. In the suggested EDTL-PCGRCC method, a wiener filtering (WF) based noise removal technique takes place for noise removal procedure. Furthermore, the EDTL-PCGRCC method uses an improved MobileNetV2 technique to derive the feature vector from pre-processed images. Furthermore, the classification of RCC takes place using the Elman Neural Network (ENN) mechanism. Lastly, improved artificial ecosystem optimization (IAEO) is applied for the parameter selection of the ENN model. The efficiency of the EDTL-PCGRCC method is assessed under the biomedical image dataset. The empirical findings reported the robustness of the EDTL-PCGRCC method under various measures.