Assessment of Lung Cancer by Pathologists using CT Scans with Deep Learning Methods: A Review

Lung cancer remains a leading cause of cancer-related mortality worldwide, with early and
accurate diagnosis being critical for improving survival outcomes. Histopathological examination of
biopsy tissue continues to be the gold standard for definitive diagnosis, offering insights into tumor type
and molecular markers. However, with the increasing reliance on imaging— particularly computed
tomography (CT) scans—for lung cancer screening, staging, and monitoring, there is a growing need for
automated, robust, and scalable diagnostic support tools. Deep learning, particularly convolutional neural
networks (CNNs), has emerged as a powerful technique for the automated analysis of CT images,
enabling accurate detection, segmentation, and classification of lung nodules. This review consolidates
current advancements in the application of deep learning models to CT imaging, covering key
methodologies, segmentation architectures (e.g., U-Net, V-Net), classification frameworks (e.g., 3D
CNNs, ResNet), and radiogenomic approaches that bridge imaging features with molecular data. Key
public datasets such as LIDC-IDRI and NSCLC-Radiomics have been instrumental in training and
validating these models. The integration of AI into clinical workflows has demonstrated the potential to
augment pathologists’ capabilities by providing rapid, consistent, and quantitative assessments.
Nonetheless, challenges persist regarding model interpretability, data variability, and regulatory approval
for clinical deployment. This review emphasizes the synergistic role of CT-based deep learning and
histopathological evaluation in achieving precise and personalized lung cancer diagnosis.