Transformer-assisted convolutional feature extraction with deep representation learning models for lung and colon cancer diagnosis using histopathological images.

Cancer is one of the deadliest diseases that can be caused by some metabolic anomalies or a convergence of transmitted disorders. Lung and colon cancer (LCC) are posited as the most common causes of death and disability in the modern world. Identification of the tumour in an initial stage before it spreads further inside the body will lessen the chance of death. Histopathological images (HIs) are extensively applied by medical professionals for identification, and they are highly essential in predicting patients' chances of survival. Usually, detection cancer using HIs requires a lengthy expert evaluation, but advanced technology allows for faster and more efficient diagnosis. Recently, artificial intelligence (AI) and DL methods are prevalently utilized for quick inspection, decision making and effectual handling of high-dimensional data, such as multi-dimensional anatomical images and videos. In this manuscript, a Lung and Colon Cancer Diagnosis via Transformer-Assisted Convolutional Feature Extraction and Deep Representation Learning (LCCD-TCFEDRL) technique using HI analysis is proposed. The aim is to develop an effective diagnostic model for LCC by utilizing advanced analytical methods to improve early detection accuracy and support improved treatment outcomes. Initially, the guided image filtering (GIF) model is employed in the image pre-processing stage to enhance the quality of images by eliminating the noise. Furthermore, the CoAtNet method is utilized for feature extraction to recognize and isolate the most relevant information from raw data. Finally, the bidirectional temporal convolutional network (BiTCN) with Adan optimizer (AO) is employed for the LCC classification process. The experimentation of the LCCD-TCFEDRL methodology is examined under the LCC HIs dataset. The comparison study of the LCCD-TCFEDRL methodology portrayed a superior accuracy value of 99.36% over existing models.