Impacted teeth with a high prevalence rate pose a significant challenge in dental diagnosis and treatment planning. Traditional methods heavily rely on manual assessment, which can be time-consuming and prone to human error for classifying the signs of anomalies. This research proposes a novel approach for impacted teeth analysis using a custom Convolution Neural Network (CNN) architecture. Panoramic radiographs are employed for this research as they cover the whole dental arch, from individual teeth to the maxilla and mandible. It is significantly influenced by the diverse angles and shapes of teeth with non-similar patterns of individuals’ impacted teeth. The proposed CNN architecture leverages the power of deep learning to classify impacted teeth from panoramic radiographs automatically. This study evaluated the performance of pre-trained models and YOLO architectures for the teeth dataset. The custom CNN architecture outperforms these models and shows an testing accuracy of 95.99%. Overall, our research showcases the effectiveness of deep learning in revolutionizing the analysis of impacted teeth, paving the way for more efficient and personalized treatment strategies in dentistry.