Leveraging Deep Object Detection Models for Early Detection of Cancerous Lung Nodules in Chest X-Rays

Timely identification of lung cancer is of utmost importance due to its high fatality rate, making it imperative for effective treatment strategies. The identification of malignant lung nodules from Chest radiographs is a frequently neglected domain due to the limited capability of X-ray imaging in capturing minute entities. This study employed a variety of deep-learning models to identify malignant nodules in chest X-ray images. The study employed three deep-learning models, specifically FatserRCNN, Yolov5, and EfficientDet, for the purpose of training and validating their ability to detect lung nodules in X-ray images. Among the various models examined, YOLOv5 demonstrated the highest efficacy in the detection of malignant nodules. The model attained precision, recall, and mean average precision (mAP) scores of 89%, 84.6%, and 83% respectively when evaluated on the NODE21 dataset. When compared to notable works, we have attained the highest recall. In addition, a web application was developed to enable users to access real-time detection outcomes utilizing the YOLOv5 model that has been trained. In conclusion, the findings of our study highlight the potential of utilizing deep learning techniques to enhance the precision and effectiveness of cancer detection in the field of medical imaging.