Machine learning approaches for diagnosing depression: A narrative review

Background and aims: Over ۳۰۰ million individuals worldwide are at risk for depression, makingit one of the most important public health problems. However, behavioral diagnostic techniquescontinue to make it difficult to make a clinical diagnosis of depression. However, the StatisticalManual of Mental Disorders (DSM-V) and doctors’ individual judgement were still usedin the clinical diagnosis of depression. Due to the lack of reliable laboratory diagnostic criteria,accurate depression identification and diagnosis remained challenging. The purpose of this narrativereview is to describe and demonstrate the impact of using artificial intelligent in diagnosisof depression.Method: we report studies that investigated risk factors of postpartum depression by searchingthe database, Scopus, PubMed, Science Direct, Up-to-date, ProQuest in the period ۲۰۱۰-۲۰۲۲published articles about the factors associated with postpartum depression were assessed in Farsiand English. The search strategy included a combination of keywords include depression, artificialintelligent, diagnosis, machine learning and psychiatric disorder.Conclusion: Literature review showed that Different machine learning techniques, such as logisticregression, the random forest, the support vector machine and/or the artificial neural networkin the case of numerical data, and the random forest in the case of genomic data, would be appropriatefor different types of data for the early diagnosis of depression. Their reported performancemetrics for accuracy and AUC ranged from ۶۰.۱ to ۱۰۰.۰ and ۶۴.۰ to ۹۶.۰, respectively.An efficient, non-invasive decision support system for the detection of depression is provided viamachine learning.

کلیدواژه ها:

artificial intelligence ، depression ، machine learning ، medicine