Interpreting Multiclass Lung Cancer from CT Scans using Grad-CAM on Lightweight CNN Layers

Lung cancer is one of the most prevalent and deadly cancer kinds. CT scan imaging has made it easier to observe in-depth at different lung cancer conditions. Convolutional Neural Networks have already been utilized for detecting objects in images. This research implements a lightweight CNN architecture in classifying two types of cancerous images and distinguishing them from healthy ones. AlexNet, ResNet101 V2, DenseNet201, MobileNet V2, and other sizable deep learning and transfer learning models were tested against this unique architecture with global max pooling. The proposed CNN model overlaps in performance with much bigger models whilst using about eight times fewer parameters, thus being efficient. Apart from making the architecture smaller, Grad-CAM was used to explain the convolutional flow over different convolution layers. This integration of the Grad-CAM technique into the analysis adds a crucial layer of transparency to the deep-learning models. By vi-sualizing the regions within lung cancer images that significantly influence predictions, we aim to enhance the interpretability of our model, fostering trust and better understanding among healthcare professionals.