ConvNeXtV2-Enhanced YOLOv8 Model Detects Dental Caries on Panoramic Radiographs with 62.6% AP50

An AI model combining ConvNeXtV2 and YOLOv8 with genetic algorithm tuning achieved 78.4% precision, 53.6% recall, and 62.6% AP50 for detecting dental caries on panoramic radiographs, outperforming baseline YOLOv8.

• The model integrates ConvNeXtV2 blocks into YOLOv8's C2f modules and uses a genetic algorithm for hyperparameter optimization.
• On a dataset of 474 annotated panoramic radiographs, the model achieved a mean average precision (AP50) of 62.6%, with a recall of 53.6%.
• The study's authors note that recall remains a limitation, indicating room for improvement in sensitivity.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.