Deep Learning in Thoracic Oncology: Meta-Analytical Insights into Lung Nodule Early-Detection Technologies

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Accurate lung nodule detection on CT scans is vital for diagnosing thoracic cancers.
• Deep learning, especially CNNs, offers automated detection potential.
• This study systematically reviews AI model accuracy for lung nodule detection.

Purpose of the Study:

• To evaluate the diagnostic accuracy of deep learning models for lung nodule detection.
• To determine lesion-wise sensitivity as the primary performance metric.
• To analyze factors influencing model performance.

Main Methods:

• Systematic review and meta-analysis of 48 studies up to November 2023.
• Random-effects model used for pooled diagnostic performance assessment.
• CLAIM and QUADAS-2 tools ensured methodological rigor.

Main Results:

• Pooled sensitivity was 79% (independent datasets) and 85% (all datasets).
• Performance variability linked to dataset characteristics and study methods.
• Factors like demographics and data splitting influenced results.

Conclusions:

• Deep learning models show significant potential for lung nodule detection.
• Need for more diverse datasets and standardized evaluation protocols.
• Further research required to improve generalizability and clinical applicability.