Clinical Report: FAD-YOLO: An Efficient Framework for Distinguishing AIS from MIA in Pulmonary CT Imaging

Overview

FAD-YOLO is a novel detection framework designed to distinguish between adenocarcinoma in situ (AIS) and minimally invasive adenocarcinoma (MIA) in pulmonary CT imaging. It demonstrates high precision and recall rates while maintaining a lightweight architecture.

Background

Lung cancer remains the leading cause of cancer-related mortality globally, with adenocarcinoma being the most prevalent subtype. Accurate differentiation between AIS and MIA is crucial for surgical planning. Current automated detection methods face challenges due to the subtle imaging characteristics of these lesions.

Data Highlights

Key Findings

• FAD-YOLO achieved a precision of 93.8% and recall of 93.4% on an internal test set.
• The model demonstrated an mAP@50 of 93.6% and mAP@50–95 of 70.8%.
• FAD-YOLO reduced parameters by 18.7% compared to the baseline YOLO12n.
• On an external test set, FAD-YOLO achieved mAP@50 of 91.7% without fine-tuning.

Clinical Implications

The FAD-YOLO framework may assist radiologists in accurately distinguishing between AIS and MIA.

Conclusion

FAD-YOLO presents an advancement in automated detection of AIS and MIA in pulmonary CT imaging, balancing accuracy and efficiency.

Related Resources & Content

1. Behera et al., 2022 -- Meta-analysis on disease-free survival rates for AIS and MIA
2. IARC Publications Website, 2021 -- WHO Classification of Thoracic Tumours
3. Automated Assessment of Aortic Contrast-Enhanced CT Angiograms for Tailored Dose Optimization in Patients
4. European Radiology — Assessing the Role of AI-Enhanced 3D Gradient Echo Imaging in the Rapid and Effective Detection of Pulmonary Nodules
5. European Radiology — Detection of Contrast-Enhanced Breast Lesions in Rapid Screening MRI Utilizing Deep Learning Techniques
6. European Radiology — Increasing pulmonary artery visibility and diagnostic confidence with ultra-high resolution photon-counting detector CT pulmonary angiography
7. Automated Assessment of Aortic Contrast-Enhanced CT Angiograms
8. Assessing the Role of AI-Enhanced 3D Gradient Echo Imaging
9. Detection of Contrast-Enhanced Breast Lesions in Rapid Screening MRI
10. IARC Publications Website - Thoracic Tumours
11. Frontiers | CT-based habitat radiomics for preoperative differentiation of adenocarcinoma in situ/minimally invasive adenocarcinoma from invasive adenocarcinoma manifesting as ground-glass nodules: a multicenter study
12. Clinical Outcomes of Ground-Glass Nodules Detected in a CT Lung Cancer Screening Program