Early Diagnosis of Lung Adenocarcinoma Leveraging YOLO Architectures

Lung adenocarcinoma, a primary subtype of Non-Small Cell Lung Cancer (NSCLC), poses a significant health threat, given its high mortality rate. Early detection remains crucial but challenging due to limitations in current screening techniques. This study explores the efficacy of leveraging deep learning, specifically the YOLO algorithm, to enhance the accuracy of adenocarcinoma detection from medical imaging. Using data from The Cancer Imaging Archive (TCIA), the study focused on CT and PET scans to train the YOLOv8s and YOLOv10s models. The findings indicate that YOLOv10s outperforms YOLOv8s, achieving a mean average precision (mAP50) of 0.732, thus demonstrating superior accuracy in identifying adenocarcinoma. Despite limitations such as dataset scope, the models exhibited proficiency in discrimination tasks as evidenced by confusion matrices and precision-recall metrics. Results highlight the potential for employing YOLO in clinical settings for real-time, accurate detection of adenocarcinoma, with implications for improving diagnostic efficiency and patient out- comes. Future work will focus on testing additional cancer types and expanding dataset diversity to enhance model robustness.