According to the most recent data from the WHO, cancer remains a leading global health concern, causing approximately 10 million deaths each year. Lung cancer stands out as a major contributor, responsible for an estimated 2.21 million fatalities in 2020 alone. Given the rising prevalence of lung cancer, early diagnosis is crucial to improving patient outcomes and ensuring timely treatment. This study utilizes histopathological images obtained from the microscopic examination of tissue biopsies to differentiate between various types of lung cancer. The focus is on identifying distinct Histological Growth Patterns (such as Acinar, any Papillary, any Solid, and Lepidic) for each cancer subtype and predicting disease severity scores through the application of Advanced Signal Processing Techniques. Following this, the deep learning model EfficientNetB7 is employed to classify three categories: Adenocarcinoma, Benign, and Squamous Cell Carcinoma. The results provide clinicians with the ability to accurately distinguish between benign and malignant cases, allowing for more personalized treatment strategies. The model's performance is rigorously evaluated using metrics such as Classification Accuracy, F1-Score, Precision, and Recall. The proposed method marks a significant improvement in classification accuracy, increasing from 97.5% to 99.8%, thereby outperforming existing methodologies.