Feasibility of deep learning for dental caries classification in bitewing radiographs based on the ICCMS™ radiographic scoring system

Diagnostic performance of artificial intelligence-aided caries detection on bitewing radiographs: a systematic review and meta-analysis

“…Despite demonstrating good SP in overall caries detection and conforming with the trend of declining diagnostic performance in caries classification tasks, these findings deviate markedly from the average performance of the other studies included in the meta-analyses for caries classification ( Fig. 5 ) [29] , [30] . Valuable insights can be gained from examining the potential causes behind this outcome.…”

“…Furthermore, data variability and possible class imbalances, as manifested by the prevalence of caries of different extensions in the training dataset, directly impact performance metrics. This is particularly relevant since Chen et al, 2022 report a caries prevalence of 24.8 % (in a total of 978 images) in the dataset in comparison to 41.7 % (in a total of 2758 images) reported by Panyarak et al, 2023 [30] . Finally, discrepancies in performance can be directly attributed to variations in AI model capabilities.…”

“…The final database search yielded 935 unique citations. Of these, 62 potential references were identified, and 14 studies were included in the qualitative assessment ( supplementary materials Appendix IV) [19] , [20] , [21] , [22] , [23] , [24] , [25] , [26] , [27] , [28] , [29] , [30] , [31] , [32] . There were several reasons for study exclusion as listed in Appendix V. An overview of the study flow and eligibility assessment process is summarized in Fig.…”

“…Through author communications, we received the unreported contingency table data for three studies. One of which was successfully included in the meta-analyses for overall, dentin, and enamel caries detection [30] . However, it was not possible to include the other two studies.…”

Transforming Dental Caries Diagnosis Through Artificial Intelligence-Based Techniques

“…The application of deep learning techniques for dental caries classification, built on the foundation of the ICDAS™ scoring system, holds substantial promise. Such an approach can significantly enhance both the precision and the speed of dental caries diagnosis [ 34 ].…”