Publications

Diagnosis of COVID-19 at point of care is vital to the containment of the global pandemic. Point of care ultrasound (POCUS) provides rapid imagery of lungs to detect COVID-19 in patients in a repeatable and cost effective way. Previous work has used public datasets of POCUS videos to train an AI model for diagnosis that obtains high sensitivity. Due to the high stakes application we propose the use of robust and explainable techniques. We demonstrate experimentally that robust models have more stable predictions and offer improved interpretability. A framework of contrastive explanations based on adversarial perturbations is used to explain model predictions that aligns with human visual perception.

Estimations of human joint torques can provide quantitative, clinically valuable information to inform patient care, plan therapy, and assess the design of wearable robotic devices. Standard methods for estimating joint torques are limited to laboratory or clinical settings since they require expensive equipment to measure joint kinematics and ground reaction forces. Wearable sensor data combined with neural networks may offer a less expensive and obtrusive estimation method.We present a method of mapping joint torque estimates obtained from motion capture and ground reaction forces to wearable sensor data. We use several different neural networks to learn the torque mapping for the ankle joints during standing, walking, running, and sprinting. Our results show that neural networks that consider time (recurrent and long short-term memory networks) outperform feedforward network architectures, producing results in the range of 0.005-0.008 N m/kg mean squared error (MSE) when compared to the inverse dynamics model on which it was trained. As a point of reference, the typical measurement errors from inverse dynamics models are in the range of 0.0004-0.0064 N m/kg MSE. Errors tended to increase with locomotion speed, with the highest errors during sprinting and the lowest during standing or walking. Future work may investigate model generalizability across sensor placements, subjects, locomotion variants, and usage duration. The proposed method relies on learning from a motion capture dataset, but once the model is built, the method uses wearable sensors that enable torque estimation without the motion capture data. These methods also have potential uses for the design and testing of wearable robotic systems outside of a laboratory environment.

Compared to other major medical imaging modalities such as X-ray, computed tomography (CT), and magnetic resonance imaging, medical ultrasound (US) has unique attributes that make it the preferred modality for many clinical applications. In particular, US is nonionizing, portable, and provides real-time imaging, with adequate spatial and depth resolution to visualize tissue dynamics. The ability to measure Doppler information is also important, particularly for measuring blood flows. The small size of US transducers is a key attribute for intravascular applications. In addition, accessibility has been increased with the use of portable US, which continues to move toward a smaller footprint and lower cost. Nowadays, some US probes can even be directly connected to a phone or tablet. On the other hand, US also has unique challenges, particularly in that image quality is highly dependent on the operator’s skill in acquiring images based on the proper position, orientation, and probe pressure. Additional challenges that further require operator skill include the presence of noise, artifacts, limited field of view, difficulty in imaging structures behind bone and air, and device variability across manufacturers. Sonographers become highly proficient through extensive training and long experience, but high intra- and interobserver variability remains. This skill dependence has limited the wider use of US by healthcare providers who are not US imaging specialists. Recent advances in machine learning (ML) have been increasingly applied to medical US (Brattain, Telfer, Dhyani, Grajo, & Samir, 2018), with a goal of reducing intra- and interobserver variability as well as interpretation time. As progress toward these goals is made, US use by nonspecialists is expected to proliferate, including nurses at the bedside or medics in the field. The acceleration in ML applications for medical US can be seen from the increasing number of publications (Fig. 8.1) and Food and Drug Administration (FDA) approvals (Table 8.1) in the past few years. Fig. 8.1 shows that cardiovascular applications (spanning the heart, brain and vessels) have received the most attention, compared to other organs. Table 8.1 shows that pace of US FDA-cleared artificial intelligence (AI) products that combine AI and ultrasound is accelerating. Of note, many of the products have been approved over the last couple of years. Companies such as Butterfly Network (Guilford, CT) have also demonstrated AI-driven applications for portable ultrasound and more FDA clearances are expected to be published. The goals of this chapter are to highlight the recent progress, as well as the current challenges and future opportunities. Specifically, this chapter addresses topics such as the following: (1) what is the current state of machine learning for medical US application, both in research and commercially; (2) what applications are receiving the most attention and have performance improvements been quantified; (3) how do ML solutions fit in an overall workflow; and (4) what open-source datasets are available for the broader community to contribute to progress in this field. The focus is on cardiovascular applications (Section Cardiovascular/echocardiography), but common themes and differences for other applications for medical US are also summarized (Section Breast, liver, and thyroid ultrasound). A discussion is offered in Discussion and outlook section.

The capability to avoid other air traffic is a fundamental component of the layered conflict management system to ensure safe and efficient operations in the National Airspace System. The evaluation of systems designed to mitigate the risk of midair collisions of manned aircraft are based on large-scale modeling and simulation efforts and a quantitative volume defined as a near midair collision (NMAC). Since midair collisions are difficult to observe in simulation and are inherently rare events, basing evaluations on NMAC enables a more robust statistical analysis. However, an NMAC and its underlying assumptions for assessing close encounters with manned aircraft do not adequately consider the different characteristics of smaller UAS-only encounters. The primary contribution of this paper is to explore quantitative criteria to use when simulating two or more smaller UASs in sufficiently close proximity that a midair collision might reasonably occur and without any mitigations to reduce the likelihood of a midair collision. The criteria assumes a historically motivated upper bound for the collision likelihood and subsequently identify the smallest possible NMAC analogs. We also demonstrate the NMAC analogs can be used to support modeling and simulation activities.

In this paper we study dynamical distortion problems in future electrical energy systems with high renewable penetration. We introduce a new time-domain modeling of electrical energy systems comprising inverter-controlled distributed energy resources (DERs). This modeling is first used to quantify the relations between distortions and real/reactive power dynamics. Next, to ensure acceptable Quality of Service (QoS), a novel nonlinear distributed inverter control is introduced. Sufficient conditions are established for the guaranteed performance of the proposed control. These conditions further support the practical implementation of the derived controller. The effectiveness of this enhanced control is illustrated using simulations for the case of avoiding system instability during sudden grid reconfigurations. Simulations also show that distortions can be suppressed in systems with parallel-connected solar photovoltaics (PVs).

Lunar polar regions and permanently shadowed regions (PSRs) are a key component of NASA's exploration objectives for the lunar surface, given their potential for a high abundance of volatiles like water. The Massachusetts Institute of Technology (MIT) Big Idea Challenge Team proposed the use of deployable towers to support robotic and remote exploration of these PSRs, alleviating limitations imposed by the rugged terrain. This deployable tower technology (called MELLTT) could enable an extended ecosystem on the lunar surface. This paper seeks to build on this initial concept by showcasing the advantages of self-deploying lightweight lunar towers through the development of various payload concepts. The payloads include 5-kg packages for an initial proof-of-concept deployment, as well as 50-kg payloads and payloads across multiple towers for future exploration architectures. The primary goal of a 5-kg tower payload is to return unique scientific data from a PSR while minimizing risk to a tower technology demonstration mission. Concepts include passive imagers to provide a step-change improvement in resolution, solar reflectors capable of illuminating PSRs, communications infrastructure for human and robotic exploration, a power beaming demonstration, and a PSR impactor. These payloads demonstrate the utility of towers on the lunar surface and how incremental improvements in the capability of towers can further NASA's exploration program.

To support integration of unmanned aerial systems into the airspace, the low altitude airspace needs to be characterized. Identifying the frequency of different aircraft types, such as rotorcraft or fixed wing single engine, given criteria such as altitude, airspace class, or quantity of seats can inform surveillance requirements, flight test campaigns, or simulation safety thresholds for detect and avoid systems. We leveraged observations of Automatic Dependent Surveillance-Broadcast (ADS-B) equipped aircraft by the OpenSky Network for this characterization.

The US Army envisions heterogeneous teams of advanced machines and humans that will collaborate together to achieve a common mission goal. It is essential for commanders to quickly and effectively respond to dynamic mission environments with agile re-tasking and computerized aids for plan definition/redefinition, and to perform some tasks with bounded autonomy. Workload constraints limit an individual's ability to concurrently control many platforms, so some mission segments many need to be autonomous or to be quickly selected via a tactics playbook. Denied environments also dictate the need for machine participants in some mission segments to be autonomous (or semi-autonomous). A Multi-Agent System (MAS) provides a natural paradigm for describing a system of agents that work together in such environments. An agent can be a human or machine, but is generally a machine. Creating MAS systems and requirements has proved to be a formidable task due to mission complexities, the necessity to deal with unforeseen circumstances, and the general difficulty of defining autonomous behaviors. We define a process called Multi-Agent Systems Collaborative Teaming (MASCOT) Definition Process that starts with a Subject Matter Experts (SME), produces a set of agent specifications, and derives system requirements in sufficient detail to define a MAS that can be modeled in a test-bed, used for facilitation of a safety analysis, and produced into an actual system. The MASCOT process also enables concurrent development of an effects based ontology. We demonstrate the MASCOT process on an example case study to show the efficacy of our process.

In 2019, the U.S. National Institute of Standards and Technology (NIST) conducted a leaderboard style speaker recognition challenge using conversational telephone speech (CTS) data extracted from the unexposed portion of the Call My Net 2 (CMN2) corpus previously used in the 2018 Speaker Recognition Evaluation (SRE). The SRE19 CTS Challenge was organized in a similar manner to SRE18, except it offered only the open training condition. In addition, similar to the NIST i-vector challenge, the evaluation set consisted of two subsets: a progress subset, and a test subset. The progress subset comprised 30% of the trials and was used to monitor progress on the leaderboad, while the remaining 70% of the trials formed the test subset, which was used to generate the official final results determined at the end of the challenge. Which subset (i.e., progress or test) a trial belonged to was unknown to challenge participants, and each system submission had to contain outputs for all of trials. The CTS Challenge also served as a prerequisite for entrance to the main SRE19 whose primary task was audio-visual person recognition. A total of 67 organizations (forming 51 teams) from academia and industry participated in the CTS Challenge and submitted 1347 valid system outputs. This paper presents an overview of the evaluation and several analyses of system performance for all primary conditions in the CTS Challenge. Compared to the CTS track of the SRE18, the SRE19 CTS Challenge results indicate remarkable improvements in performance which are mainly attributed to 1) the availability of large amounts of in-domain development data from a large number of labeled speakers, 2) speaker representations (aka embeddings) extracted using extended and more complex end-to-end neural network frameworks, and 3) effective use of the provided large development set.

In 2019, the U.S. National Institute of Standards and Technology (NIST) conducted the most recent in an ongoing series of speaker recognition evaluations (SRE). There were two components to SRE19: 1) a leaderboard style Challenge using unexposed conversational telephone speech (CTS) data from the Call My Net 2 (CMN2) corpus, and 2) an Audio-Visual (AV) evaluation using video material extracted from the unexposed portions of the Video Annotation for Speech Technologies (VAST) corpus. This paper presents an overview of the Audio-Visual SRE19 activity including the task, the performance metric, data, and the evaluation protocol, results and system performance analyses. The Audio-Visual SRE19 was organized in a similar manner to the audio from video (AfV) track in SRE18, except it offered only the open training condition. In addition, instead of extracting and releasing only the AfV data, unexposed multimedia data from the VAST corpus was used to support the Audio-Visual SRE19. It featured two core evaluation tracks, namely audio only and audio-visual, as well as an optional visual only track. A total of 26 organizations (forming 14 teams) from academia and industry participated in the Audio-Visual SRE19 and submitted 102 valid system outputs. Evaluation results indicate: 1) notable performance improvements for the audio only speaker recognition task on the challenging amateur online video domain due to the use of more complex neural network architectures (e.g., ResNet) along with soft margin losses, 2) state-of-the-art speaker and face recognition technologies provide comparable person recognition performance on the amateur online video domain, and 3) audio-visual fusion results in remarkable performance gains (greater than 85% relative) over the audio only or visual only systems.