Lung cancer is the cancer that spreads the fastest and is typically detected at an advanced stage. It may cause death with late diagnosing and improper treatment. A computer-aided detection method is required to categorise the lung nodule with the greatest degree of accuracy in order to avoid delays in diagnosis due to advancements in medical imaging methods like computed tomography (CT) scans. This study proposed a novel architecture D3DR_MKCA based on Deep Residual network incorporating convolutional block attention module (CBAM) which applied on different scale feature maps to classify lung nodules. CBAM improves the representation power of Residual Network. Initially lung nodules are efficiently segmented with the help of Location Aware Encoding Network and those segmented nodules are further classified into Adenocarcinoma, Small Cell Carcinoma, Large Cell Carcinoma, and Squamous Cell Carcinoma cancerous tissues with the help of proposed D3DR_MKCA deep architecture. A Large-Scale CT and PET/CT Dataset for Lung Cancer Diagnosis (Lung-PET-CT-Dx)are used for performance analysis and the D3DR_MKCA model archives F1-score up to 90.96%.