Lesion and caries dedection in dental x-rays with YOLOv7 algorithm

The healthcare sector has made significant advancements in achieving more consistent diagnoses and successful outcomes through the utilization of deep learning algorithms. In the context of oral health, dental inflammations over time can lead to the development of cavities in teeth. When care is not taken for dental care, bacteria form in the person's mouth over time. This bacteria combine with food to damage tooth enamel and tooth decay occurs. Tooth decay is an important health problem that negatively affects our lives. Until recently inflamed teeth are immediately removed to prevent damage to other areas. Nowadays, with the development of imaging methods, various treatments have been used to prevent tooth lose and the success rate has increased. The diagnosis of these cavities and lesions are usually diagnosed by taking dental X-ray images. However, even when performed by expert professionals, such diagnoses may exhibit variations. Hence, there arises a need for leveraging deep learning-based approaches. This study aims to automatically detect lesions and cavities in dental X-ray images using a YOLOv7-based deep learning algorithm. Four hundred panoramic dental X-ray images, meticulously annotated by an expert dentist, were employed for this purpose. This annotation ensured precise delineation of lesion and cavity regions. Subsequently, these images were transformed into YOLOv7 format using the Roboflow platform and utilized for model training. The YOLOv7 algorithm was trained on this dataset and demonstrated a high level of accuracy in identifying lesions and cavities in dental X-ray images. The performance of the model was evaluated using performance metrics such as precision, recall and mAP resulting in an average precison(mAP) of 38%. The obtained results have the potential to enhance diagnostic consistency among healthcare institutions, minimize personal interpretation errors and optimize treatment plans.