CAREER: Autonomous Wearable Computing for Personalized Healthcare

CAREER: Autonomous Wearable Computing for Personalized Healthcare CAREER: Autonomous Wearable Computing for Personalized Healthcare Wearables are poised to transform health and wellness through automation of cost-effective, objective, continuous, and real-time health monitoring and interventions. Currently, however, computational models for these systems are designed based on labeled training data collected in controlled settings. This has created a number of real impediments to scalability of wearables because (i) collecting sufficiently large amounts of labeled sensor data is a time consuming, labor-intensive, and expensive pro-cess that has been identified as a major barrier to personalized and precision medicine; and (ii) wearables are deployed in highly dynamic environments of the end-users whose physical, behavioral, social, and environ-mental context undergoes consistent changes. Such changes result in drastic performance degradation of the computational models trained in laboratory settings. The overarching goal of this CAREER proposal is to develop algorithms and tools that enable future wearables to learn in-situ autonomously, operate in-the-wild reliably, and adapt to the changing context of their users automatically. Intellectual Merit. This project is transformative because it will yield the first end-to-end solutions that will enable autonomous, reliable, and robust learning of wearable computing algorithms in naturalistic settings without need for collecting any labeled sensor data. The proposal aims to develop foundations of computa-tional autonomy for wearable-based health monitoring and interventions through the following intellectually rigorous objectives: Autonomous sensor data labeling: we will research methods of autonomous labeling of sensor data in a new setting (i.e., target), based on labeled training data collected in a different setting (i.e., source). We will devise algorithms for cross-subject, cross-context, cross-platform, and cross-modality labeling by developing combinatorial approaches to identify an optimal mapping from target to source. Reliable label inference: we will develop algorithms to refine labels learned through autonomous sensor data labeling or directly received from another source, to reliably infer target labels from uncertain and sporadic knowledge of a potentially unreliable and heterogeneous source. We will design adaptive non-parametric label propagation techniques that combine noisy labels with local observations of target to make reliable inferences. Robust model generation: we will investigate methods of training computational models that are robust to unknown parameters of source and adaptive to dynamically changing signal attributes. We will develop structured detection algorithms that take into account interdependency of physical/physiological events and known attributes of the application to accurately detect a wide range of health events. Validation in-the-wild: in addition to extensive in-lab testing, we will evaluate the proposed algorithms through in-field trials involving patients with heart failure, cardiac arrhythmia, and cancer. Broader Impacts. The various thrusts of this research will not only address the technical challenges in de-veloping highly performant and autonomous wearable systems, but will also enable actual monitoring of a variety of populations. This project has major broader impacts on conducting high-precision chronic disease management and on the availability of wearable-based consumer applications. In the medium-to-long-term, this will lead to the development of products and business around the concept of computational autonomy and its use in automation of health management and countless other yet un-envisioned applications. The PIs educational plan will focus on developing a new ambassador program to increase interest of under-represented minority community college students in STEM careers in general and in computer science and engineering careers in particular, developing a novel patron program to improve retention of transferred underrepresented minority students through student and parental exposure to wearable health research, en-gaging undergraduate students in research, and establishment of an interdisciplinary research-based curricu-lum on computational autonomy. The PI is committed to provide mentoring to graduate students as well as minorities; he is currently advising seven graduate students including four female students. Keywords: Wearable Computing, Self-Configuration, Machine Learning, Uncontrolled Environments.