SISTEM DETEKSI LOKASI KARIES GIGI PADA CITRA GAMBAR RONTGEN MENGGUNAKAN YOLO V8 DENGAN VISUALISASI AREA TERDETEKSI

Dental caries is one of the most common oral health problems, particularly among children and adolescents. According to the 2023 Indonesian Health Survey (SKI), the prevalence of caries among Indonesians aged over 3 years remains high at 82.8%, with 84.8% of cases occurring in children aged 5–9 years. Early detection of caries is often challenging due to limitations in conventional methods, such as visual examination and radiography, which heavily rely on the expertise and accuracy of dental professionals. To address this issue, this study developed an automatic dental caries detection system based on periapical radiographic images using the YOLOv8 (You Only Look Once version 8) algorithm. The dataset used consisted of 254 annotated periapical dental X-ray images labeled by professional dentists. The images were obtained from Mulya Dental Clinic and the Roboflow dataset, with further annotation performed through the Roboflow platform.
The model training process was conducted using a GPU on Google Colab, configured with 300 epochs and a batch size of 16. The dataset was divided into three parts: training (70%), validation (20%), and testing (10%). Evaluation results showed a recall value of 0.77358 and a precision value of 0.71141, indicating that the model demonstrated good sensitivity in detecting caries while being fairly accurate in distinguishing between carious and non-carious areas. This system was integrated into a web-based application using Streamlit to facilitate its practical use by both medical professionals and patients. It is expected that this system can support a more accurate and efficient diagnostic process, reducing the risk of human error caused by fatigue or oversight.
Keywords: Dental Caries, YOLOv8, Detection, Periapical Radiograph, Deep Learning