IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE, VOL. 14, NO. 2, MARCH 2010 241 Experimental Analysis of a Mobile Health System for Mood Disorders Tammara Massey, Member, IEEE, Gustavo Marfia, Miodrag Potkonjak, Member, IEEE, and Majid Sarrafzadeh, Fellow, IEEE Abstract—Depression is one of the leading causes of disabil- ity. Methods are needed to quantitatively classify emotions in or- der to better understand and treat mood disorders. This research proposes techniques to improve communication in body sensor network (BSN) that gathers data on the affective states of the patient. These BSNs can continuously monitor, discretely quan- tify, and classify a patient’s depressive states. In addition, data on the patient’s lifestyle can be correlated with his/her physiological conditions to identify how various stimuli trigger symptoms. This continuous stream of data is an improvement over a snapshot of localized symptoms that a doctor often collects during a medical examination. Our research first quantifies how the body interferes with communication in a BSN and detects a pattern between the line of sight of an embedded device and its reception rate. Then, a mathematical model of the data using linear programming tech- niques determines the optimal placement and number of sensors in a BSN to improve communication. Experimental results show that the optimal placement of embedded devices can reduce power cost up to 27% and reduce hardware costs up to 47%. This research brings researchers a step closer to continuous, real-time systemic monitoring that will allow one to analyze the dynamic human phys- iology and understand, diagnosis, and treat mood disorders. Index Terms—Affective computing, body sensor networks (BSN), embedded systems, health informatics, medical applications. I. INTRODUCTION T HE WORLD Health Organization (WHO) estimates that depressive disorders affect approximately 121 million adults and is one of the leading causes of disability in the world [1]. Depression is a chronic mental disorder, where peo- ple are usually despondent, lose interest in activities, feel guilty, have feelings of low self-worth, lack energy, lose their appetite, and have trouble concentrating. Bipolar disorders are a subset of depressive disorders, where patients swing between moods of mania and depression. This disorder is also called manic Manuscript received November 19, 2008; revised July 13, 2009. First pub- lished October 23, 2009; current version published March 17, 2010. This work was supported in part by Microsoft Research. T. Massey is with the Johns Hopkins University/Applied Physics Laboratory, Laurel, MD 20723 USA (e-mail: tammara.massey@jhuapl.edu). G. Marfia is with the Dipartimento di Scienze dell’Informazione, Universit` a di Bologna, Bologna, 40126 Italy (e-mail: marfia@cs.unibo.it). M. Potkonjak and M. Sarrafzadeh are with the Department of Computer Sci- ence, University of California, Los Angeles, CA 90095 USA (e-mail: miodrag@ cs.ucla.edu; majid@cs.ucla.edu). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TITB.2009.2034738 depression and the periods of mania are characterized by a sub- stantial increase in energy, insomnia, racing thoughts, feelings of grandiose, short attention spans, and anxious feelings. Depression and other mood disorders are difficult disorders to diagnose, understand, and treat due to difficulties in classifying emotion and the fluctuating nature of the disease. Individuals with manic depression continuously have changes in energy, mood, and activity that may not be apparent in office visits. A quantitative method to continuously track patients and pro- vide more data on their moods and emotions could improve how doctors and healthcare clinicians diagnosis and treat the disease. Recent research has explored the use of noninvasive mobile sensors that can be placed on the human body to measure the vi- tal signs of a patient. These body sensor networks (BSN) allow patients to be continuously monitored remotely and are useful for many medical applications, including manic depression. Pre- vious work by Pentland and coworkers has demonstrated that accelerometer sensors measuring body movements can classify depression states and track the treatment of patients [2]. These sensors can aid in the recognition, interpretation, or inference of human emotion. The ability to classify depression states and emotion can fundamentally improve the treatment of patients with manic depression and mood disorders by being able to detect patterns in the patient and suggest interventions. However, an open challenge is how to create a communi- cation infrastructure that will allow patients to continuously communicate affective cognitive states through the use of wear- able sensors. BSN can consist of wearable systems embedded in cloth or portable embedded systems that can be carried by the patient in a similar manner as the cell phone. In this paper, challenges in connectivity of BSN and interference from the human body itself is addressed. BSN have different properties than most traditional wireless sensor networks (WSN). A new class of distributed embedded systems BSN, is rapidly evolving and the need to develop archi- tectures from experimental data is needed [3]. Our approach will build upon our growing understanding and experience with wire- less ad hoc sensor networks. WSN share key features with BSN. WSN and BSN sense, self-configure their resources, and provide actuation under severe communication and memory constraints. In addition, the systems are tightly coupled to the physical world and must adapt and self-organize without human intervention. Due to the dynamic environment that BSN are deployed in, re- alistic data based on actual events are essential. However, BSNs differ significantly from WSN in that the system must deal with a large amount of interference from the body, the systems are 1089-7771/$26.00 © 2009 IEEE