Publications

Hemorrhage is the leading cause of preventable death from trauma. Traditionally, vital signs have been used to detect blood loss and possible hemorrhagic shock. However, vital signs are not sensitive for early detection because of physiological mechanisms that compensate for blood loss. As an alternative, machine learning algorithms that operate on an arterial blood pressure (ABP) waveform acquired via photoplethysmography have been shown to provide an effective early indicator. However, these machine learning approaches lack physiological interpretability. In this paper, we evaluate the importance of nine ABP-derived features that provide physiological insight, using a database of 40 human subjects from a lower-body negative pressure model of progressive central hypovolemia. One feature was found to be considerably more important than any other. That feature, the half-rise to dicrotic notch (HRDN), measures an approximate time delay between the ABP ejected and reflected wave components. This delay is an indication of compensatory mechanisms such as reduced arterial compliance and vasoconstriction. For a scale of 0% to 100%, with 100% representing normovolemia and 0% representing decompensation, linear regression of the HRDN feature results in root-mean-squared error of 16.9%, R2 of 0.72, and an area under the receiver operating curve for detecting decompensation of 0.88. These results are comparable to previously reported results from the more complex black box machine learning models. Clinical Relevance- A single physiologically interpretable feature measured from an arterial blood pressure waveform is shown to be effective in monitoring for blood loss and impending hemorrhagic shock based on data from a human lower-body negative pressure model of progressive central hypolemia.

Lung ultrasound (LUS) as a diagnostic tool is gaining support for its role in the diagnosis and management of COVID-19 and a number of other lung pathologies. B-lines are a predominant feature in COVID-19, however LUS requires a skilled clinician to interpret findings. To facilitate the interpretation, our main objective was to develop automated methods to classify B-lines as pathologic vs. normal. We developed transfer learning models based on ResNet networks to classify B-lines as pathologic (at least 3 B-lines per lung field) vs. normal using COVID-19 LUS data. Assessment of B-line severity on a 0-4 multi-class scale was also explored. For binary B-line classification, at the frame-level, all ResNet models pretrained with ImageNet yielded higher performance than the baseline nonpretrained ResNet-18. Pretrained ResNet-18 has the best Equal Error Rate (EER) of 9.1% vs the baseline of 11.9%. At the clip-level, all pretrained network models resulted in better Cohen's kappa agreement (linear-weighted) and clip score accuracy, with the pretrained ResNet-18 having the best Cohen's kappa of 0.815 [95% CI: 0.804-0.826], and ResNet-101 the best clip scoring accuracy of 93.6%. Similar results were shown for multi-class scoring, where pretrained network models outperformed the baseline model. A class activation map is also presented to guide clinicians in interpreting LUS findings. Future work aims to further improve the multi-class assessment for severity of B-lines with a more diverse LUS dataset.

Humans need to work in many types of extreme environments where there is a need to stay safe and even to improve performance. Examples include: medical providers treating infectious disease, people responding to other biological or chemical hazards, firefighters, astronauts, pilots, divers, and people working outdoors in extreme hot or cold temperatures. Wearable technology is ubiquitous in the consumer market but is still needed for extreme environments. For these applications, it is particularly challenging to meet requirements to be actionable, accurate, acceptable, integratable, and affordable. To provide insight into these needs and possible solutions and the technology trade-offs involved, several examples are provided. A physiological monitoring example is described for predicting and avoiding heat injury. A cognitive monitoring example is described for estimating cognitive workload, with broader applicability to a variety of conditions, such as cognitive fatigue and depression. Finally, eye tracking is considered as a promising wearable sensing modality with applications for both physiological and cognitive monitoring. Concluding thoughts are offered on the compelling need for wearable technology in the face of pandemics, wildfires, and climate change, but also for global projects that can uplift mankind, such as long-duration spaceflight and missions to Mars.

Early detection of diseases such as COVID-19 could be a critical tool in reducing disease transmission by helping individuals recognize when they should self-isolate, seek testing, and obtain early medical intervention. Consumer wearable devices that continuously measure physiological metrics hold promise as tools for early illness detection. We gathered daily questionnaire data and physiological data using a consumer wearable (Oura Ring) from 63,153 participants, of whom 704 self-reported possible COVID-19 disease. We selected 73 of these 704 participants with reliable confirmation of COVID-19 by PCR testing and high-quality physiological data for algorithm training to identify onset of COVID-19 using machine learning classification. The algorithm identified COVID-19 an average of 2.75 days before participants sought diagnostic testing with a sensitivity of 82% and specificity of 63%. The receiving operating characteristic (ROC) area under the curve (AUC) was 0.819 (95% CI [0.809, 0.830]). Including continuous temperature yielded an AUC 4.9% higher than without this feature. For further validation, we obtained SARS CoV-2 antibody in a subset of participants and identified 10 additional participants who self-reported COVID-19 disease with antibody confirmation. The algorithm had an overall ROC AUC of 0.819 (95% CI [0.809, 0.830]), with a sensitivity of 90% and specificity of 80% in these additional participants. Finally, we observed substantial variation in accuracy based on age and biological sex. Findings highlight the importance of including temperature assessment, using continuous physiological features for alignment, and including diverse populations in algorithm development to optimize accuracy in COVID-19 detection from wearables.

Objective Exertional heat stroke (EHS), characterised by a high core body temperature (Tcr) and central nervous system (CNS) dysfunction, is a concern for athletes, workers and military personnel who must train and perform in hot environments. The objective of this study was to determine whether algorithms that estimate Tcr from heart rate and gait instability from a trunk-worn sensor system can forward predict EHS onset. Methods Heart rate and three-axis accelerometry data were collected from chest-worn sensors from 1806 US military personnel participating in timed 4/5-mile runs, and loaded marches of 7 and 12 miles; in total, 3422 high EHS-risk training datasets were available for analysis. Six soldiers were diagnosed with heat stroke and all had rectal temperatures of >41°C when first measured and were exhibiting CNS dysfunction. Estimated core temperature (ECTemp) was computed from sequential measures of heart rate. Gait instability was computed from three-axis accelerometry using features of pattern dispersion and autocorrelation. Results The six soldiers who experienced heat stroke were among the hottest compared with the other soldiers in the respective training events with ECTemps ranging from 39.2°C to 40.8°C. Combining ECTemp and gait instability measures successfully identified all six EHS casualties at least 3.5 min in advance of collapse while falsely identifying 6.1% (209 total false positives) examples where exertional heat illness symptoms were neither observed nor reported. No false-negative cases were noted. Conclusion The combination of two algorithms that estimate Tcr and ataxic gate appears promising for real-time alerting of impending EHS.

Hemorrhage is a leading cause of trauma death, particularly in prehospital environments when evacuation is delayed. Obtaining central vascular access to a deep artery or vein is important for administration of emergency drugs and analgesics, and rapid replacement of blood volume, as well as invasive sensing and emerging life-saving interventions. However, central access is normally performed by highly experienced critical care physicians in a hospital setting. We developed a handheld AI-enabled interventional device, AI-GUIDE (Artificial Intelligence Guided Ultrasound Interventional Device), capable of directing users with no ultrasound or interventional expertise to catheterize a deep blood vessel, with an initial focus on the femoral vein. AI-GUIDE integrates with widely available commercial portable ultrasound systems and guides a user in ultrasound probe localization, venous puncture-point localization, and needle insertion. The system performs vascular puncture robotically and incorporates a preloaded guidewire to facilitate the Seldinger technique of catheter insertion. Results from tissue-mimicking phantom and porcine studies under normotensive and hypotensive conditions provide evidence of the technique's robustness, with key performance metrics in a live porcine model including: a mean time to acquire femoral vein insertion point of 53 plus or minus 36 s (5 users with varying experience, in 20 trials), a total time to insert catheter of 80 plus or minus 30 s (1 user, in 6 trials), and a mean number of 1.1 (normotensive, 39 trials) and 1.3 (hypotensive, 55 trials) needle insertion attempts (1 user). These performance metrics in a porcine model are consistent with those for experienced medical providers performing central vascular access on humans in a hospital.

Parkinson's disease (PD) is a chronic movement disorder that produces a variety of characteristic movement abnormalities. The ubiquity of wrist-worn accelerometry suggests a possible sensor modality for early detection of PD symptoms and subsequent tracking of PD symptom severity. As an initial proof of concept for this technological approach, we analyzed the U.K. Biobank data set, consisting of one week of wrist-worn accelerometry from a population with a PD primary diagnosis and an age-matched healthy control population. Measures of movement dispersion were extracted from automatically segmented gait data, and measures of movement dimensionality were extracted from automatically segmented low-movement data. Using machine learning classifiers applied to one week of data, PD was detected with an area under the curve (AUC) of 0.69 on gait data, AUC = 0.84 on low-movement data, and AUC = 0.85 on a fusion of both activities. It was also found that classification accuracy steadily improved across the one-week data collection, suggesting that higher accuracy could be achievable from a longer data collection. These results suggest the viability of using a low-cost and easy-to-use activity sensor for detecting movement abnormalities due to PD and motivate further research on early PD detection and tracking of PD symptom severity.

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 purpose of this study was to develop an automated method for classifying liver fibrosis stage >=F2 based on ultrasound shear wave elastography (SWE) and to assess the system's performance in comparison with a reference manual approach. The reference approach consists of manually selecting a region of interest from each of eight or more SWE images, computing the mean tissue stiffness within each of the regions of interest and computing a resulting stiffness value as the median of the means. The 527-subject database consisted of 5526 SWE images and pathologist-scored biopsies, with data collected from a single system at a single site. The automated method integrates three modules that assess SWE image quality, select a region of interest from each SWE measurement and perform machine learning-based, multi-image SWE classification for fibrosis stage >=F2. Several classification methods were developed and tested using fivefold cross-validation with training, validation and test sets partitioned by subject. Performance metrics were area under receiver operating characteristic curve (AUROC), specificity at 95% sensitivity and number of SWE images required. The final automated method yielded an AUROC of 0.93 (95% confidence interval: 0.90-0.94) versus 0.69 (95% confidence interval: 0.65-0.72) for the reference method, 71% specificity with 95% sensitivity versus 5% and four images per decision versus eight or more. In conclusion, the automated method reported in this study significantly improved the accuracy for >=F2 classification of SWE measurements as well as reduced the number of measurements needed, which has the potential to reduce clinical workflow.

Breathing rate was estimated from chest-worn accelerometry collected from 1,522 servicemembers during training by a wearable physiological monitor. A total of 29,189 hours of training and sleep data were analyzed. The primary purpose of the monitor was to assess thermal-work strain and avoid heat injuries. The monitor design was thus not optimized to estimate breathing rate. Since breathing rate cannot be accurately estimated during periods of high activity, a qualifier was applied to identify sedentary time periods, totaling 8,867 hours. Breathing rate was estimated for a total of 4,179 hours, or 14% of the total collection and 47% of the sedentary total, primarily during periods of sleep. The breathing rate estimation method was compared to an FDA 510(K)-cleared criterion breathing rate sensor (Zephyr, Annapolis MD, USA) in a controlled laboratory experiment, which showed good agreement between the two techniques. Contributions of this paper are to: 1) provide the first analysis of accelerometry-derived breathing rate on free-living data including periods of high activity as well as sleep, along with a qualifier that effectively identifies sedentary periods appropriate for estimating breathing rate; 2) test breathing rate estimation on a data set with a total duration that is more than 60 times longer than that of the largest previously reported study, 3) test breathing rate estimation on data from a physiological monitor that has not been expressly designed for that purpose.