Machine Learning Based on Transfer Learning for Predicting Lung and Colon Cancer
TL;DR
The proposed system may support early detection, which in turn may reduce mortality rates and improve health care outcomes, and the fact that the use of DL and ML has great promise in health care, in particular in middle-and low-income countries, which have limited access to advanced diagnostic tools is put forth.
Abstract
Lung and colon (LC) cancer is the leading cause of death from cancer worldwide. Early detection is key in this disease, which at the same time proves to be a challenge because, in its early stages, symptoms are few. Present a method that offers a comprehensive approach using machine learning (ML) and deep learning (DL) for the early detection of LC cancer, achieved through analysis of histopathology images (HPI). Used modified transfer learning (TL) models, which we augmented with average-pooling 2D layers for better feature extraction from the HPI dataset. Then, feed these features into ML classifiers to evaluate their performance. Achieved very impressive results across the board, with the Efficient Net B2 and the SVC classifier reporting an accuracy, recall, F1 score, and precision of 0.9996, and an AUC of 0.9998, which outperforms present methods. Also, we put forth the fact that the use of DL and ML has great promise in health care, in particular in middle-and low-income countries, which have limited access to advanced diagnostic tools. Our proposed system may support early detection, which in turn may reduce mortality rates and improve health care outcomes.