Depresi merupakan salah satu masalah kesehatan mental yang umum dan menjadi penyebab utama disabilitas di seluruh dunia. Deteksi dini sangat penting karena gejala awal sering tidak dikenali atau diabaikan. Penelitian ini mengusulkan sistem deteksi dini depresi berbasis kecerdasan buatan melalui analisis citra wajah menggunakan pendekatan hibrida Convolutional Neural Network (CNN). Data diperoleh dari rekaman wawancara klinis pasien dengan diagnosis depresi dan individu normal. Ekstraksi fitur dilakukan menggunakan tiga arsitektur CNN, yaitu VGG16, ResNet101V2, dan MobileNetV3, yang kemudian digabungkan untuk membentuk representasi fitur gabungan. Selanjutnya, algoritma Random Forest digunakan sebagai klasifikator. Hasil eksperimen menunjukkan bahwa kombinasi VGG16 dan MobileNetV3 memberikan performa terbaik dengan akurasi 94.55%, presisi 98.04%, dan spesifisitas 98.18%. Temuan ini menunjukkan bahwa pendekatan hibrida CNN dengan Random Forest berpotensi menjadi metode efektif untuk mendukung deteksi dini depresi berbasis citra wajah. Abstract Depression is a major mental health disorder and one of the leading causes of disability worldwide. Early detection is crucial since initial symptoms are often overlooked or unrecognized. This study proposes an artificial intelligence–based early detection system for depression through facial image analysis using a hybrid Convolutional Neural Network (CNN) approach. Facial data were obtained from clinical interview recordings of patients diagnosed with depression and healthy individuals. Feature extraction was carried out using three CNN architectures VGG16, ResNet101V2, and MobileNetV3 whose features were then combined into a joint representation. A Random Forest algorithm was employed for classification. Experimental results indicate that the hybrid model combining VGG16 and MobileNetV3 achieved the best performance, with 94.55% accuracy, 98.04% precision, and 98.18% specificity. These findings demonstrate that the proposed CNN Random Forest hybrid approach is a promising method for supporting early depression detection based on facial features.
English
Depresi merupakan salah satu masalah kesehatan mental yang umum dan menjadi penyebab utama disabilitas di seluruh dunia. Deteksi dini sangat penting karena gejala awal sering tidak dikenali atau diabaikan. Penelitian ini mengusulkan sistem deteksi dini depresi berbasis kecerdasan buatan melalui analisis citra wajah menggunakan pendekatan hibrida Convolutional Neural Network (CNN). Data diperoleh dari rekaman wawancara klinis pasien dengan diagnosis depresi dan individu normal. Ekstraksi fitur dilakukan menggunakan tiga arsitektur CNN, yaitu VGG16, ResNet101V2, dan MobileNetV3, yang kemudian digabungkan untuk membentuk representasi fitur gabungan. Selanjutnya, algoritma Random Forest digunakan sebagai klasifikator. Hasil eksperimen menunjukkan bahwa kombinasi VGG16 dan MobileNetV3 memberikan performa terbaik dengan akurasi 94.55%, presisi 98.04%, dan spesifisitas 98.18%. Temuan ini menunjukkan bahwa pendekatan hibrida CNN dengan Random Forest berpotensi menjadi metode efektif untuk mendukung deteksi dini depresi berbasis citra wajah. Abstract Depression is a major mental health disorder and one of the leading causes of disability worldwide. Early detection is crucial since initial symptoms are often overlooked or unrecognized. This study proposes an artificial intelligence–based early detection system for depression through facial image analysis using a hybrid Convolutional Neural Network (CNN) approach. Facial data were obtained from clinical interview recordings of patients diagnosed with depression and healthy individuals. Feature extraction was carried out using three CNN architectures VGG16, ResNet101V2, and MobileNetV3 whose features were then combined into a joint representation. A Random Forest algorithm was employed for classification. Experimental results indicate that the hybrid model combining VGG16 and MobileNetV3 achieved the best performance, with 94.55% accuracy, 98.04% precision, and 98.18% specificity. These findings demonstrate that the proposed CNN Random Forest hybrid approach is a promising method for supporting early depression detection based on facial features.
Affiliations come from OpenAlex; journal OAI-PMH metadata carries none. Authors without an OpenAlex ID are matched by name, which can merge or split people.