Assessment of Bipolar Disorder Using Heterogeneous Data of Smartphone-Based Digital Phenotyping

Abstract

In mental health disorder, Bipolar Disorder (BD) is one of the most common mental illness. Using rating scales for assessment is one of the approaches for diagnosing and tracking BD patients. However, the requirement for manpower and time is heavy in the process of evaluation. In order to reduce the cost of social and medical resources, this study collects the user's data by the App on smartphones, consisting of location data (GPS), self-report scales, daily mood, sleeping time and records of multi-media (text, speech, video) which are heterogeneous digital phenotyping data, to build a database. The features of each heterogeneous digital phenotyping data are extracted independently. Lasso Regression and ElasticNet Regression methods are employed to predict the score of Hamilton Depression Rating Scale (HAM-D) and Young Mania Rating Scale (YMRS), as a reference for the evaluation of BD. As incomplete and missing data are very common in medical research, the ensemble method is adopted to combine the results from different models trained with different combinations of missing data. The collected heterogeneous digital phenotyping data from 84 BD patients were used for training and evaluation of the proposed approach based on five-fold cross validation method. Experimental results show that the performance of the assessment system using the proposed method are encouraging.

Machine Learning for Multimodal Mental Health Detection: A Systematic Review of Passive Sensing Approaches

“…The more popular ML algorithms for supervised learning are linear regression, logistic regression, and support vector machines (SVMs). Based on comparisons conducted in existing studies, stochastic gradient descent [ 43 ] and least absolute shrinkage and selection operator (lasso) regression [ 200 , 213 ] models performed the best in respective investigations on different feature combinations, i.e., the prior on audio, visual and textual features and the latter on wearable sensor signals, but these models are yet to be compared under similar settings. In addition to the traditional or linear algorithms mentioned above, the following subsection discusses a subset of supervised learning approaches utilizing neural networks.…”

“…Depression AV [43, SM [20,25,98, SS [99,100,104,105, WS [149][150][151][155][156][157][158]164,169,[171][172][173][178][179][180][181][182] Suicidal intent AV [100,183,184] SM [185][186][187][188][189] SS [100,147,190] WS [181,182,191] Bipolar disorder AV [101][102][103][192][193][194][195][196][197][198][199][200] SM…”

Digital Phenotyping for Mood Disorders: Methodology-Oriented Pilot Feasibility Study

“…Recognizing the potential clinical value of passive monitoring of affective disorders and closer tracking of depressive symptoms, and standardizing how mobile health studies are structured and analyzed are necessary to advance this growing field [8,38]. Hence, in these preliminary findings, we provide a comprehensive description of the methodology of our study assessing the feasibility of using a mobile health app in depression, a description of the technology used for the study, and the factors and barriers that need to be considered in iteratively developing a future study design in this domain.…”

“…In relation, this research needs to be understood in the context of any potential biases. While there have been prior studies using digital phenotyping methods for mood and bipolar disorder [7,8], the focus has often been the resulting behavior features and not an assessment of which types of people agree to partake in this research and the quality of the data gathered from their phones. Both factors matter in the interpretation of later results.…”

Automatic Bipolar Disorder Assessment Using Machine Learning With Smartphone-Based Digital Phenotyping

“…These models have a higher number of parameters involved in calculations and can more efficiently consider time-series relationships to achieve long-term tracking and favorable outcomes. Reliance on nonlinear models for prediction may lead to excessive errors due to overfitting [48].…”