Havish S. Kantheti

Introduction. Depression is a mood disorder that causes a persistent feeling of sadness and loss of interest.¹ Depression is prevalent illness that impacts the quality of life of nearly 10% globally.^{1, 2} The World Health Organization (WHO) estimates depression will be the leading cause of disease burden by 2030.¹ Phenotypical similarities within major depressive disorder (MDD) led to delayed proper diagnosis.² Although multiple treatment options are available, treatment is done on a trial and error basis, increasing the time to effective treatment.³ Machine learning (ML) techniques have gained traction as a way to diagnose and predict treatment outcomes in MDD.^2,3 ML can be described as methods which learn from data to make classification of new data inputs.² Application to MDD allows predictions of future occurrences and progression of the disease sate based on defined attributes.^2,3 The purpose of this study is to investigate current applications of ML to MDD diagnosis and treatment. Methods. One work utilized smart watch sensor data to collect activity and light exposure measurements in order to predict the onset of depression.⁴ Another study examined the using EEG as an input into various ML model types to use it as an indicator for depression.⁵ Two other works tested machine learning systems to predict the response to serotonin reuptake inhibitor (SSRI) based on factors such as patient demographics, validated biomarkers, and EEG measurements.^6,7 Results. Using light smart watch inputs and logit ML model a precision of 0.929 and an accuracy of 0.910 were achieved.⁴ EEG data as a feature input for depression prediction showed that ensemble support vector machine (SVM) obtained the best accuracy at ~89%.⁵ Biomarker and demographics input showed that a random forests model could be used to predict treatment outcomes at a specificity ranging from 0.63 – 0.85.⁶ EEG baseline and two week measurements inputted to a SVM model obtained an accuracy of 82.4% for predicting escitalopram outcomes. Conclusions. Studies have shown that machine learning can be applied to both diagnosis and treatment prediction of MDD at a reasonably high accuracy to help address some of the current short comings in the field. Their utilization in the psychiatric clinical workflow could potentially help improve quality of care of some patients.

1. Malhi, Gin S, and J John Mann. “Depression.” The Lancet, vol. 392, no. 10161, 2018, pp. 2299–2312., doi:10.1016/s0140-6736(18)31948-2.
2. Gao, Shuang, et al. “Machine Learning in Major Depression: From Classification to Treatment Outcome Prediction.” CNS Neuroscience & Therapeutics, vol. 24, no. 11, 2018, pp. 1037–1052., doi:10.1111/cns.13048.
3. Shatte, Adrian B., et al. “Machine Learning in Mental Health: a Scoping Review of Methods and Applications.” Psychological Medicine, vol. 49, no. 09, 2019, pp. 1426–1448., doi:10.1017/s0033291719000151.
4. Kim H, Lee SH, Lee SE, Hong S, Kang HJ, Kim N. Depression Prediction by Using Ecological Momentary Assessment, Actiwatch Data, and Machine Learning: Observational Study on Older Adults Living Alone. JMIR mHealth and uHealth. 2019;7(10). doi:10.2196/14149
5. Li X, Zhang X, Zhu J, et al. Depression recognition using machine learning methods with different feature generation strategies. Artificial Intelligence in Medicine. 2019;99:101696. doi:10.1016/j.artmed.2019.07.004
6. Athreya AP, Neavin D, Carrillo-Roa T, et al. Pharmacogenomics-Driven Prediction of Antidepressant Treatment Outcomes: A Machine-Learning Approach With Multi-trial Replication. Clinical Pharmacology & Therapeutics. 2019;106(4):855-865. doi:10.1002/cpt.1482
7. Zhdanov A, Atluri S, Wong W, et al. Use of Machine Learning for Predicting Escitalopram Treatment Outcome From Electroencephalography Recordings in Adult Patients With Depression. JAMA Network Open. 2020;3(1). doi:10.1001/jamanetworkopen.2019.18377