Accurate and early detection of lung cancer is needed to save the lives of patients. After COVID– 19, lung cancer contributes more to death worldwide. To eradicate this, we propose a deep learning-based diagnostic framework that integrates Integer Wavelet Transform (IWT) with a customized Fused ResNet50 architecture to classify lung CT images into three categories: Normal cases, Bengin cases, and Malignant cases. Initially, the input CT images are preprocessed and decomposed using IWT to extract LL (approximation) and LH (horizontal detail) Sub-Bands, which capture essential spatial-frequency characteristics of lung tissues. These Sub-Bands are resized and fused channel-wise to form a two-channel composite image, which serves as input to the network. The proposed model modifies the conventional ResNet50 by introducing parallel convolutional pathways for each Sub-Band in the early layers, followed by concatenation and joint feature learning in the deeper layer. This fusion strategy effectively combines low-frequency and edge related features, enhancing the model’s ability to distinguish between subtle texture variations in different lung cancer types. The model is trained and evaluated on the publicly available The IQ-OTHNCCD lung cancer dataset using an 80:20 train-test split. The experimental results show that Fused ResNet50 achieves a superior classification performance, attaining an overall accuracy of 99%, precision of 88%, recall of 97%, and F1 score of 99%. The architecture also demonstrates robust convergence and reduced overfitting due to the Sub-Band fusion mechanism. This approach presents a reliable and interpretable solution for automated lung cancer detection using CT images.