Predicting Risk of Antenatal Depression and Anxiety Using Multi-Layer Perceptrons and Support Vector Machines.

Area of Science:

• Medical Informatics
• Psychiatry
• Machine Learning

Background:

• Perinatal depression and anxiety are significant mental health issues affecting mothers and infants, often going undetected.
• High prevalence rates, particularly in low-income countries, underscore the need for effective screening tools.
• Underdiagnosis can lead to adverse outcomes, including impaired mother-child bonding and severe maternal mental health crises.

Purpose of the Study:

• To develop and evaluate a machine learning model for predicting the risk of antenatal depression and anxiety in pregnant women.
• To address the challenge of underdiagnosis and under-resourcing in perinatal mental healthcare.
• To provide a screening tool that can be integrated into routine obstetric care.

Main Methods:

• A multi-layer perceptron neural network (MLP-NN) classifier was proposed for risk prediction.
• ReliefF algorithm was employed for feature selection prior to model training.
• The system was trained and validated on a dataset of 500 Pakistani women during their antenatal period.

Main Results:

• The MLP-NN model demonstrated strong performance in predicting antenatal depression and anxiety.
• The area under the receiver operating characteristic curve (AUC) reached 88% for depression and 85% for anxiety using the MLP classifier.
• Support vector classifiers also showed competitive performance, with AUCs of 80% and 77% for depression and anxiety, respectively.

Conclusions:

• The proposed MLP-NN based system shows significant potential for screening perinatal depression and anxiety.
• This tool can serve as a valuable facilitator for early identification during routine gynecological and obstetrics visits.
• Implementing such systems can help mitigate the negative impacts of undiagnosed perinatal mental health conditions on mothers and infants.