FusionDxNet: A Fusion-Based Approach for Robust Lung & Colon Cancer Classification Using Histopathological Images

Cancer is still one among the most significant global health challenges, with lung and olorectal cancers ranking among the most commonly diagnosed types. A prompt and precise diagnosis is essential for achieving effective treatment outcomes, yet traditional diagnostic techniques are often time-consuming and reliant on expert interpretation. In this study, we introduce FusionDxNet, an advanced deep learning- based framework designed for the automated classification of lung and colon cancers from histopathological images. The proposed model leverages the strengths of two powerful pretrained convolutional neural networks—ResNet50 and MobileNet—for robust feature extraction. The extracted features are fused and passed through a hy- brid architecture comprising Logistic Regression, a custom 1D CNN, and an ensemble classifier to deliver the final prediction. This multi-stage fusion approach effectively captures complex patterns across different image types, enhancing model generaliz- ability and classification performance. Extensive experiments conducted on publicly available datasets demonstrate that FusionDxNet achieves impressive results, with 99.90% accuracy for colon cancer, 99.17% for lung cancer, and 99.38% for multi-class classification involving five categories. These results significantly outperform several existing state-of-the-art techniques. The findings suggest that FusionDxNet offers a highly reliable and efficient tool for aiding pathologists in early and precise cancer detection. Its implementation could streamline diagnostic workflows and contribute to better clinical outcomes through faster and more accurate decision-making.