Joint Modeling of Heterogeneous Sensing Data for Depression Assessment via Multi-task Learning

Abstract

Depression is a common mood disorder that causes severe medical problems and interferes negatively with daily life. Identifying human behavior patterns that are predictive or indicative of depressive disorder is important. Clinical diagnosis of depression relies on costly clinician assessment using survey instruments which may not objectively reflect the fluctuation of daily behavior. Self-administered surveys, such as the Quick Inventory of Depressive Symptomatology (QIDS) commonly used to monitor depression, may show disparities from clinical decision. Smartphones provide easy access to many behavioral parameters, and Fitbit wrist bands are becoming another important tool to assess variables such as heart rates and sleep efficiency that are complementary to smartphone sensors. However, data used to identify depression indicators have been limited to a single platform either iPhone, or Android, or Fitbit alone due to the variation in their methods of data collection. The present work represents a large-scale effort to collect and integrate data from mobile phones, wearable devices, and self reports in depression analysis by designing a new machine learning approach. This approach constructs sparse mappings from sensing variables collected by various tools to two separate targets: self-reported QIDS scores and clinical assessment of depression severity. We propose a so-called heterogeneous multi-task feature learning method that jointly builds inference models for related tasks but of different types including classification and regression tasks. The proposed method was evaluated using data collected from 103 college students and could predict the QIDS score with an R 2 reaching 0.44 and depression severity with an F1-score as high as 0.77. By imposing appropriate regularizers, our approach identified strong depression indicators such as time staying at home and total time asleep.

Using behavioral rhythms and multi-task learning to predict fine-grained symptoms of schizophrenia

“…Specifically, the authors predicted the well-being for a subgroup of people who shared similar personality traits and behaviors by treating the prediction for these people as different tasks. A more recent work by Lu et al 29 developed a MTL method to jointly model sensing data collected from different smartphone platforms (Android and iOS) for depression detection, where predicting self-reported assessment scores and clinical severity of depression on each platform were considered four different tasks. Taken together, the aim of this paper is to predict more fine-grained symptom trajectories of schizophrenia in terms of patients' self-reported EMA scores using clinically meaningful rhythm features-the different types of cyclic patterns in patients' behaviors and their surrounding environment that are extracted from their passive mobile sensor data.…”

Smartphone-based Mental State Estimation: A Survey from a Machine Learning Perspective

“…Feature LOC-7, the living area size, represents an imaginary circle encompassing the various locations that a user traveled across on a particular day [21]. Canzian et al Travel distance [21], [32], [35], [37], [41], [42], [43], [44], [45], [46], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61]…”

“…Time spent at significant location (home, clinic, office, school, etc.) [32], [39], [41], [42], [43], [47], [48], [49], [53], [55], [56], [58], [59], [60], [61], [62], [63], [64], [65], [66] LOC -3 Number of places visited [32], [37], [39], [41], [42], [48], [53], [55], [56], [57], [60], [61], [67] LOC -4 Transition time [32], [37], [41], [42], [43], [50], [61], [64], [66] LOC -5 Routine index [32], [37], [41], [42], [60] LOC-6…”

“…The stress and anxiety category includes studies that used PSS, GAD-7, SIAS, and STAI. The depression category includes studies that used PHQ-2, PHQ-8, PHQ-9, CES-D, HDRS-17, HAMD, QID, [21], [45], [48], [51], [52], [54], [62], [64], [65] [45], [51], [62], [65], [68], [69] [21], [48], [52], [54], [62], [65], [68], [69], [77] [69], [77] [45], [51], [62], [68] [45], [51] (100%) Depression 16 (32.0%) 10 (20.0%) 10 (20.0%) 2 (4.0%) 9 (18.0%) 3 (6.0%) 50 [32], [35], [39], [41], [42], [43], [44], [45], [47], [51], [53], [55], [61], [62], [63], [67] [39], [43], [45], [47], [51],…”

“…As a future research direction, we recommend considering a combination of smartphone and wearable devices for mental state estimation. For example, Wang et al [39] and Lu et al [61] used Fitbit to collect heart rate in addition to data collection from smartphone. Note that several surveys have examined mental healthcare and wearable devices [10], [11], [12].…”