Statistical uncertainty quantification to augment clinical decision support: a first implementation in sleep medicine.

Machine learning has the potential to change the practice of medicine, particularly in areas that require pattern recognition (e.g. radiology). Although automated classification is unlikely to be perfect, few modern machine learning tools have the ability to assess their own classification confidence to recognize uncertainty that might need human review. Using automated single-channel sleep staging as a first implementation, we demonstrated that uncertainty information (as quantified using Shannon entropy) can be utilized in a "human in the loop" methodology to promote targeted review of uncertain sleep stage classifications on an epoch-by-epoch basis. Across 20 sleep studies, this feedback methodology proved capable of improving scoring agreement with the gold standard over automated scoring alone (average improvement in Cohen's Kappa of 0.28), in a fraction of the scoring time compared to full manual review (60% reduction). In summary, our uncertainty-based clinician-in-the-loop framework promotes the improvement of medical classification accuracy/confidence in a cost-effective and economically resourceful manner.

Objectively measured sleep fragmentation is associated with incident delirium in older hospitalized patients: Analysis of data collected from an randomized controlled trial.

Delirium may lead to poor outcomes in hospitalized older adults, and sleep deprivation may contribute to its pathogenesis. Thus, we sought to measure sleep duration and fragmentation using wrist-worn actigraphy in older, hospitalized patients with and without delirium, and to determine if actigraphy-based parameters could be used to predict delirium prior to clinical recognition. We conducted a secondary analysis of data from a recent, randomized clinical trial aimed at preventing inpatient delirium. Participants (n = 70) were aged ≥ 65 years admitted to an internal medicine service. Delirium was defined by the Confusion Assessment Method, or altered mental status identified by a clinician. Sleep measurements were actigraphy-based, and included total sleep time, median sleep bout duration and other measures of sleep fragmentation. We found that total sleep duration was similar between patients with (n = 17) and without (n = 53) delirium (mean 384.9 ± SD 162.7 versus mean 456.6 ± SD 135.8 min; p = .081). Mean sleep bout times were shorter in delirious versus never-delirious patients (median 6.1 [interquartile range 4.3-8.9] versus 7.9 [interquartile range 5.7-11.3] min, p = .048). Patients with delirium had more short sleep bouts (< 10 min) and fewer longer sleep bouts (> 30 min) compared with those without delirium. Increased sleep fragmentation was present prior to the clinical recognition of delirium. Overall, delirium was associated with increased sleep fragmentation detected by actigraphy, and sleep fragmentation might be useful as a biomarker for delirium prediction in the future.

Smart Electronic Eyedrop Bottle for Unobtrusive Monitoring of Glaucoma Medication Adherence.

Glaucoma, the leading cause of irreversible blindness, affects >70 million people worldwide. Lowering intraocular pressure via topical administration of eye drops is the most common first-line therapy for glaucoma. This treatment paradigm has notoriously high non-adherence rates: ranging from 30% to 80%. The advent of smart phone enabled technologies creates promise for improving eyedrop adherence. However, previous eyedrop electronic monitoring solutions had awkward medication bottle adjuncts and crude software for monitoring the administration of a drop that adversely affected their ability to foster sustainable improvements in adherence. The current work begins to address this unmet need for wireless technology by creating a "smart drop" bottle. This medication bottle is instrumented with sensing electronics that enable detection of each eyedrop administered while maintaining the shape and size of the bottle. This is achieved by a thin electronic force sensor wrapped around the bottle and underneath the label, interfaced with a thin electronic circuit underneath the bottle that allows for detection and wireless transmission to a smart-phone application. We demonstrate 100% success rate of wireless communication over 75 feet with <1% false positive and false negative rates of single drop deliveries, thus providing a viable solution for eyedrop monitoring for glaucoma patients.

CNN-Based Person Detection Using Infrared Images for Night-Time Intrusion Warning Systems.

Night-time surveillance is important for safety and security purposes. For this reason, several studies have attempted to automatically detect people intruding into restricted areas by using infrared cameras. However, detecting people from infrared CCTV (closed-circuit television) is challenging because they are usually installed in overhead locations and people only occupy small regions in the resulting image. Therefore, this study proposes an accurate and efficient method for detecting people in infrared CCTV images during the night-time. For this purpose, three different infrared image datasets were constructed; two obtained from an infrared CCTV installed on a public beach and another obtained from a forward looking infrared (FLIR) camera installed on a pedestrian bridge. Moreover, a convolution neural network (CNN)-based pixel-wise classifier for fine-grained person detection was implemented. The detection performance of the proposed method was compared against five conventional detection methods. The results demonstrate that the proposed CNN-based human detection approach outperforms conventional detection approaches in all datasets. Especially, the proposed method maintained F1 scores of above 80% in object-level detection for all datasets. By improving the performance of human detection from infrared images, we expect that this research will contribute to the safety and security of public areas during night-time.

In-Home Sleep Recordings in Military Veterans With Posttraumatic Stress Disorder Reveal Less REM and Deep Sleep <1 Hz.

Veterans with posttraumatic stress disorder (PTSD) often report suboptimal sleep quality, often described as lack of restfulness for unknown reasons. These experiences are sometimes difficult to objectively quantify in sleep lab assessments. Here, we used a streamlined sleep assessment tool to record in-home 2-channel electroencephalogram (EEG) with concurrent collection of electrodermal activity (EDA) and acceleration. Data from a single forehead channel were transformed into a whole-night spectrogram, and sleep stages were classified using a fully automated algorithm. For this study, 71 control subjects and 60 military-related PTSD subjects were analyzed for percentage of time spent in Light, Hi Deep (1-3 Hz), Lo Deep (<1 Hz), and rapid eye movement (REM) sleep stages, as well as sleep efficiency and fragmentation. The results showed a significant tendency for PTSD sleepers to spend a smaller percentage of the night in REM (p < 0.0001) and Lo Deep (p = 0.001) sleep, while spending a larger percentage of the night in Hi Deep (p < 0.0001) sleep. The percentage of combined Hi+Lo Deep sleep did not differ between groups. All sleepers usually showed EDA peaks during Lo, but not Hi, Deep sleep; however, PTSD sleepers were more likely to lack EDA peaks altogether, which usually coincided with a lack of Lo Deep sleep. Linear regressions with all subjects showed that a decreased percentage of REM sleep in PTSD sleepers was accounted for by age, prazosin, SSRIs and SNRIs (p < 0.02), while decreased Lo Deep and increased Hi Deep in the PTSD group could not be accounted for by any factor in this study (p < 0.005). Linear regression models with only the PTSD group showed that decreased REM correlated with self-reported depression, as measured with the Depression, Anxiety, and Stress Scales (DASS; p < 0.00001). DASS anxiety was associated with increased REM time (p < 0.0001). This study shows altered sleep patterns in sleepers with PTSD that can be partially accounted for by age and medication use; however, differences in deep sleep related to PTSD could not be linked to any known factor. With several medications [prazosin, selective serotonin reuptake inhibitors (SSRIs), serotonin-norepinephrine reuptake inhibitors (SNRIs); p < 0.03], as well as SSRIs were associated with less sleep efficiency (b = -3.3 ± 0.95; p = 0.0005) and more sleep fragmentation (b = -1.7 ± 0.51; p = 0.0009). Anti-psychotics were associated with less sleep efficiency (b = -4.9 ± 1.4; p = 0.0004). Sleep efficiency was negatively impacted by SSRIs, antipsychotic medications, and depression (p < 0.008). Increased sleep fragmentation was associated with SSRIs, SNRIs, and anxiety (p < 0.009), while prazosin and antipsychotic medications correlated with decreased sleep fragmentation (p < 0.05).

Melatonin and Sleep in Preventing Hospitalized Delirium: A Randomized Clinical Trial.

PURPOSE: Studies suggest that melatonin may prevent delirium, a condition of acute brain dysfunction occurring in 20%-30% of hospitalized older adults that is associated with increased morbidity and mortality. We examined the effect of melatonin on delirium prevention in hospitalized older adults while measuring sleep parameters as a possible underlying mechanism. METHODS: This was a randomized clinical trial measuring the impact of 3 mg of melatonin nightly on incident delirium and both objective and subjective sleep in inpatients age ≥65 years, admitted to internal medicine wards (non-intensive care units). Delirium incidence was measured by bedside nurses using the confusion assessment method. Objective sleep measurements (nighttime sleep duration, total sleep time per 24 hours, and sleep fragmentation as determined by average sleep bout length) were obtained via actigraphy. Subjective sleep quality was measured using the Richards Campbell Sleep Questionnaire. RESULTS: Delirium occurred in 22.2% (8/36) of subjects who received melatonin vs in 9.1% (3/33) who received placebo (P = .19). Melatonin did not significantly change objective or subjective sleep measurements. Nighttime sleep duration and total sleep time did not differ between subjects who became delirious vs those who did not, but delirious subjects had more sleep fragmentation (sleep bout length 7.0 ± 3.0 vs 9.5 ± 5.3 min; P = .03). CONCLUSIONS: Melatonin given as a nightly dose of 3 mg did not prevent delirium in non-intensive care unit hospitalized patients or improve subjective or objective sleep.

A State Space and Density Estimation Framework for Sleep Staging in Obstructive Sleep Apnea.

OBJECTIVE: Although the importance of sleep is increasingly recognized, the lack of robust and efficient algorithms hinders scalable sleep assessment in healthy persons and those with sleep disorders. Polysomnography (PSG) and visual/manual scoring remain the gold standard in sleep evaluation, but more efficient/automated systems are needed. Most previous works have demonstrated algorithms in high agreement with the gold standard in healthy/normal (HN) individuals-not those with sleep disorders. METHODS: This paper presents a statistical framework that automatically estimates whole-night sleep architecture in patients with obstructive sleep apnea (OSA)-the most common sleep disorder. Single-channel frontal electroencephalography was extracted from 65 HN/OSA sleep studies, and decomposed into 11 spectral features in 60 903 30 s sleep epochs. The algorithm leveraged kernel density estimation to generate stage-specific likelihoods, and a 5-state hidden Markov model to estimate per-night sleep architecture. RESULTS: Comparisons to full PSG expert scoring revealed the algorithm was in fair agreement with the gold standard (median Cohen's kappa = 0.53). Further, analysis revealed modest decreases in median scoring agreement as OSA severity increased from HN (kappa = 0.63) to severe (kappa = 0.47). A separate implementation on HN data from the Physionet Sleep-EDF Database resulted in a median kappa = 0.65, further indicating the algorithm's broad applicability. CONCLUSION: Results of this work indicate the proposed single-channel framework can emulate expert-level scoring of sleep architecture in OSA. SIGNIFICANCE: Algorithms constructed to more accurately model physiological variability during sleep may help advance automated sleep assessment, for practical and general use in sleep medicine.

Visualization of Whole-Night Sleep EEG From 2-Channel Mobile Recording Device Reveals Distinct Deep Sleep Stages with Differential Electrodermal Activity.

Brain activity during sleep is a powerful marker of overall health, but sleep lab testing is prohibitively expensive and only indicated for major sleep disorders. This report demonstrates that mobile 2-channel in-home electroencephalogram (EEG) recording devices provided sufficient information to detect and visualize sleep EEG. Displaying whole-night sleep EEG in a spectral display allowed for quick assessment of general sleep stability, cycle lengths, stage lengths, dominant frequencies and other indices of sleep quality. By visualizing spectral data down to 0.1 Hz, a differentiation emerged between slow-wave sleep with dominant frequency between 0.1-1 Hz or 1-3 Hz, but rarely both. Thus, we present here the new designations, Hi and Lo Deep sleep, according to the frequency range with dominant power. Simultaneously recorded electrodermal activity (EDA) was primarily associated with Lo Deep and very rarely with Hi Deep or any other stage. Therefore, Hi and Lo Deep sleep appear to be physiologically distinct states that may serve unique functions during sleep. We developed an algorithm to classify five stages (Awake, Light, Hi Deep, Lo Deep and rapid eye movement (REM)) using a Hidden Markov Model (HMM), model fitting with the expectation-maximization (EM) algorithm, and estimation of the most likely sleep state sequence by the Viterbi algorithm. The resulting automatically generated sleep hypnogram can help clinicians interpret the spectral display and help researchers computationally quantify sleep stages across participants. In conclusion, this study demonstrates the feasibility of in-home sleep EEG collection, a rapid and informative sleep report format, and novel deep sleep designations accounting for spectral and physiological differences.

Scalable Microfabrication Procedures for Adhesive-Integrated Flexible and Stretchable Electronic Sensors.

New classes of ultrathin flexible and stretchable devices have changed the way modern electronics are designed to interact with their target systems. Though more and more novel technologies surface and steer the way we think about future electronics, there exists an unmet need in regards to optimizing the fabrication procedures for these devices so that large-scale industrial translation is realistic. This article presents an unconventional approach for facile microfabrication and processing of adhesive-peeled (AP) flexible sensors. By assembling AP sensors on a weakly-adhering substrate in an inverted fashion, we demonstrate a procedure with 50% reduced end-to-end processing time that achieves greater levels of fabrication yield. The methodology is used to demonstrate the fabrication of electrical and mechanical flexible and stretchable AP sensors that are peeled-off their carrier substrates by consumer adhesives. In using this approach, we outline the manner by which adhesion is maintained and buckling is reduced for gold film processing on polydimethylsiloxane substrates. In addition, we demonstrate the compatibility of our methodology with large-scale post-processing using a roll-to-roll approach.