Can We Predict Depression From the Asymmetry of Electrodermal Activity?

Abstract

IntroductionWhat are the limitations of the methods for depression diagnosis?Though useful for semantic and billing purposes, DSM-based or depression rating scale-based approaches have limited utility for 1) determining subtypes of depression; 2) capturing variations over relatively short time periods (i.e., over the course of a day), and 3) predicting the course of the illness. Therefore we hypothesize that some types of depression may cause patients to have larger electrodermal activity (EDA) on the right than on the left palm.If validated, this may lead to an objective biomarker for the diagnosis, the prognosis, and the treatment of depression Participants 9 adults with Major Depressive Episode, undergoing Transcranial Magnetic Stimulation (TMS) Procedure  3 participants attended 36, and 6 participant attended 72 daily TMS sessions lasting 20-37 minutes each  Symptom severity was measured at the baseline and throughout the treatment (every 10 sessions) by a clinician blinded to the EDA Devices During the experiment the users wore on both palms the Q sensor, a wireless non-invasive sensor. Measures  Q sensor measures EDA, motion (actigraphy), and temperature.  Clinician collected symptom severity scales once after every 10 sessions: 28-item Hamilton Depression Rating Scale (HAM-D28); Quick Inventory of Depressive Symptoms (QIDS); Patient Health Questionnaire (PHQ-9). Data analysis  We evaluated the relationship between the two palmar EDA signals during TMS sessions on the days leading up to a depression measure. We applied a low-pass filter to each EDA raw signal (1024-point Hamming window, 3Hz cut-off frequency) to reduce the motion artifacts and the electrical noise. Then we calculated an average EDA level on each palm for every session and subtracted the left hand from the right hand mean value to obtain a mean difference (EDA R-L ).  We used the linear mixed-effect with random intercepts and slopes to assess the relationship between the mean EDA difference from the palms and depression measures using the following model: Dep_sc i = β 0i + β 1i * EDA R-Li + ε i Where:Dep_sc i -depression scale value for i-th person β 0i -i-th person intercept, β 0i = β 0 +µ 0i , and µ 0i ~ N(0,σ 0 2 ) β 1i -i-th person slope, β 1i = β 1 +µ 1i , and µ 1i ~ N(0,σ 1 2 ) ε i -i-th person error, and ε i~ N(0,σ 2 )

Detecting Deception and Suspicion in Dyadic Game Interactions

“…EDA has been widely used for various tasks such as seizure detection [44], engagement recognition during social interactions [23], analysis of EDA during sleep [56], or depression prediction based on EDA asymmetry [14].…”

Smartwatch Applications for Mental Health: A Qualitative Analysis of the Users� Perspectives

“…For mental health particularly, wearable computing has been investigated to address depression (Fedor, 2016), stress (Sano & Picard, 2013), (Picard, 2016), anxiety, borderline and bipolar disorder, and schizophrenia (Wang et al, 2016), (Cella et al, 2017), (Torous and Keshavan, 2018 In a search in iTune and Android store, Larsen et al (2016) identified applications for depression, bipolar disorder and suicide. They noticed that few apps were actually clinically relevant (35.3%, 347 out of 982), only nine (2.6%) had clinical effectiveness and only three referred to a published study.…”