Generalizability of Deep Learning Models for Caries Detection in Near-Infrared Light Transillumination Images

• Main Authors: Agnes Holtkamp, Karim Elhennawy, José E. Cejudo Grano de Oro, Joachim Krois, Sebastian Paris, Falk Schwendicke
• Format: Article
• Language: English
• Published: MDPI AG 2021-03-01
• Series: Journal of Clinical Medicine
• Subjects: artificial intelligence; caries; diagnostics; digital imaging/radiology; mathematical modeling
• Online Access: https://www.mdpi.com/2077-0383/10/5/961

Objectives: The present study aimed to train deep convolutional neural networks (CNNs) to detect caries lesions on Near-Infrared Light Transillumination (NILT) imagery obtained either in vitro or in vivo and to assess the models’ generalizability. Methods: In vitro, 226 extracted posterior permanent human teeth were mounted in a diagnostic model in a dummy head. Then, NILT images were generated (DIAGNOcam, KaVo, Biberach), and images were segmented tooth-wise. In vivo, 1319 teeth from 56 patients were obtained and segmented similarly. Proximal caries lesions were annotated pixel-wise by three experienced dentists, reviewed by a fourth dentist, and then transformed into binary labels. We trained ResNet classification models on both in vivo and in vitro datasets and used 10-fold cross-validation for estimating the performance and generalizability of the models. We used GradCAM to increase explainability. Results: The tooth-level prevalence of caries lesions was 41% in vitro and 49% in vivo, respectively. Models trained and tested on in vivo data performed significantly better (mean ± SD accuracy: 0.78 ± 0.04) than those trained and tested on in vitro data (accuracy: 0.64 ± 0.15; <i>p</i> < 0.05). When tested in vitro, the models trained in vivo showed significantly lower accuracy (0.70 ± 0.01; <i>p</i> < 0.01). Similarly, when tested in vivo, models trained in vitro showed significantly lower accuracy (0.61 ± 0.04; <i>p</i> < 0.05). In both cases, this was due to decreases in sensitivity (by −27% for models trained in vivo and −10% for models trained in vitro). Conclusions: Using in vitro setups for generating NILT imagery and training CNNs comes with low accuracy and generalizability. Clinical significance: Studies employing in vitro imagery for developing deep learning models should be critically appraised for their generalizability. Applicable deep learning models for assessing NILT imagery should be trained on in vivo data.