Gender Classification with Hand-Wrist Radiographs Using the Deep Learning Method Derin Öğrenme Yöntemi Kullanılarak El-Bilek Radyografileri ile Cinsiyet Sınıflandırma

Objective: Before dental procedures, hand-wrist radiographs are used to plan treatment time and determine skeletal maturity. This study aims to determine gender from hand-wrist radiographs using different deep-learning methods. Methods: The left hand-wrist radiographs of 1044 individuals (534 males and 510 females) were pre-processed to clarify the image and adjust the contrast. In the gender classification problem, AlexNet, VGG16 and VGG19 transfer learning methods were both used as separate classifiers, and the features taken from these methods were combined and given to the support vector machine (SVM) classifier. Results: The results revealed that image analysis and deep learning techniques provided 91.1% accuracy in gender determination. Conclusion: Hand-wrist radiographs exhibited sexual dimorphism and could be used in gender prediction.