| Abstract | Lung cancer, including both Non-Small Cell Lung Cancer (NSCLC) and Small Cell Lung Cancer
(SCLC), presents with diverse radiographic features that are crucial for accurate detection and
classification. NSCLC includes subtypes such as adenocarcinoma, which typically appears as peripheral
nodules, squamous cell carcinoma, which is often central and associated with cavitation, and large cell
carcinoma, characterized by large, poorly defined masses. SCLC, known for its rapid growth and
metastasis, often appears centrally near the bronchi and can lead to complications like superior vena
cava syndrome. Radiographically, lung cancer manifests as masses or nodules with varying borders,
atelectasis, lymphadenopathy, pleural effusion, and interstitial changes. However, challenges in
detection arise due to the overlap of lung cancer with other diseases such as tuberculosis and pneumonia,
and early-stage lesions that may be subtle and difficult to identify without advanced imaging techniques.
The application of deep learning models has significantly advanced detection, enabling more precise
segmentation of tumor regions, feature extraction for differentiating benign from malignant lesions, and
classification of cancer subtypes based on imaging patterns. Explainability tools like Grad-CAM further
enhance model transparency, helping clinicians better understand automated predictions and improve
diagnostic accuracy for lung cancer.
|
|---|
