Zeeshan Qamar; Eisa Alturki; Hamad Almutawtah; Mesfer Alajmi & Sager Al Shatti

Summary

The first mandibular pre-molars have been considered for sexual dimorphism on their manual morphometric characteristics. Though there is a debate on its reliability for correct predilection of the sexes. The aim of the research was to evaluate the performance of convolutional neural network (CNNs) in distinguishing between males and females based on the radiographic morphology of left mandibular first pre-molars on 2-D images (OPGs). The data was collected from 12,915 patients comprising of 5983 males and 6932 females respectively with an age group of 14-29.99 years. The radiographic data was analyzed by Darwin V7 software trailed by DenseNet121 CNNs was used for identification of males and females across 14-29.99 years of age groups. The performance of the CNNs was measured using the rate of accuracy, confusion matrices and ROC curves. The range for the rate of accuracy was 63.20 % - 73.4 % with mean standard deviation of 68.65 %±3.31 %. The AUC for the ROC curve analysis ranged between 0.63 for 29-29.99 years age group and 0.73 for 18-18.99 years age group. In conclusion it can be suggested that the morphometric analysis of the left mandibular first pre-molars can be used under certain situations where other anatomical characteristic features are unattainable for sexual determination. KEY WORDS: Artificial intelligence; Convolutional neural networks; Mandibular first pre-molar; Radiology; Sexual dimorphism.

How to cite this article

QAMAR, Z.; ALTURKI, E.; ALMUTAWTAH, H.; ALAJMI, M. & AL SHATTI, S. AI-powered CNNs analysis of mandibular first premolars radiographic morphology for sexual dimorphism in the south western population. Int. J. Morphol., 43(2):365- 372, 2025