Cost-Efficient Deep Learning Model for Colorectal Cancer Diagnosis: Leveraging Unsupervised Learning Composite Network

Abstract

Despite this, the mortality rate associated with CRC is among the highest of all malignancies. Even while prompt and accurate diagnosis can significantly improve patient outcomes, traditional diagnostic procedures are usually expensive, invasive, and time-consuming. This is despite the fact that rapid diagnosis can significantly improve patient outcomes. Deep learning has emerged as a potential avenue for the diagnosis of cancer, despite the prohibitive nature of the costs associated with developing robust models in terms of both computational and financial resources. In this article, the authors offer a deep learning model that is optimized for the diagnosis of colorectal cancer at a low cost. The model was trained using an unsupervised learning composite network (ULCN). Unsupervised learning is utilized by the ULCN in order to pre-process enormous amounts of unlabeled data in order to cut down on the necessary volume of labelled data as well as the associated cost. Our model was able to capitalize on the underlying structures and patterns in the data, which allowed it to perform far better than other methods available at the time on a test set of histopathological images. The unsupervised feature extraction phase significantly reduced the amount of time required for training, hence showcasing the efficiency of the model. The findings indicate that by combining unsupervised learning into deep learning systems, it may be possible to create a diagnostic tool that is more easily accessible and scalable in clinical settings.