Applying machine learning to consumer wearable data for the early detection of complications after pediatric appendectomy.

When children are discharged from the hospital after surgery, their caregivers often rely on subjective assessments (e.g., appetite, fatigue) to monitor postoperative recovery as objective assessment tools are scarce at home. Such imprecise and one-dimensional evaluations can result in unwarranted emergency department visits or delayed care. To address this gap in postoperative monitoring, we evaluated the ability of a consumer-grade wearable device, Fitbit, which records multimodal data about daily physical activity, heart rate, and sleep, in detecting abnormal recovery early in children recovering after appendectomy. One hundred and sixty-two children, ages 3-17 years old, who underwent an appendectomy (86 complicated and 76 simple cases of appendicitis) wore a Fitbit device on their wrist for 21 days postoperatively. Abnormal recovery events (i.e., abnormal symptoms or confirmed postoperative complications) that arose during this period were gathered from medical records and patient reports. Fitbit-derived measures, as well as demographic and clinical characteristics, were used to train machine learning models to retrospectively detect abnormal recovery in the two days leading up to the event for patients with complicated and simple appendicitis. A balanced random forest classifier accurately detected 83% of these abnormal recovery days in complicated appendicitis and 70% of abnormal recovery days in simple appendicitis prior to the true report of a symptom/complication. These results support the development of machine learning algorithms to predict onset of abnormal symptoms and complications in children undergoing surgery, and the use of consumer wearables as monitoring tools for early detection of postoperative events.

Applications of artificial intelligence in ovarian stimulation: a tool for improving efficiency and outcomes.

Because of the birth of the first baby after in vitro fertilization (IVF), the field of assisted reproductive technologies (ARTs) has seen significant advancements in the past 40 years. Over the last decade, the healthcare industry has increasingly adopted machine learning algorithms to improve patient care and operational efficiency. Artificial intelligence (AI) in ovarian stimulation is a burgeoning niche that is currently benefiting from increased research and investment from both the scientific and technology communities, leading to cutting-edge advancements with promise for rapid clinical integration. AI-assisted IVF is a rapidly growing area of research that can improve ovarian stimulation outcomes and efficiency by optimizing the dosage and timing of medications, streamlining the IVF process, and ultimately leading to increased standardization and better clinical outcomes. This review article aims to shed light on the latest breakthroughs in this area, discuss the role of validation and potential limitations of the technology, and examine the potential of these technologies to transform the field of assisted reproductive technologies. Integrating AI responsibly into IVF stimulation will result in higher-value clinical care with the goal of having a meaningful impact on enhancing access to more successful and efficient fertility treatments.

Causal inference indicates that poor responders have similar outcomes with the antagonist protocol compared with flare.

OBJECTIVE: To use causal inference to investigate whether the flare or antagonist protocol is better for poor responders going through controlled ovarian stimulation. DESIGN: A retrospective study. SETTING: Retrieval cycles from the Society for Assisted Reproductive Technology Clinic Outcomes Reporting System. PATIENTS: Patients in the United States underwent autologous in vitro fertilization cycles from 2014 to 2019 using either the flare or antagonist protocol. INTERVENTION: Not applicable. MAIN OUTCOME MEASURE: Primary outcomes included oocytes retrieved, fertilized oocytes (2PNs), blastocysts, the cumulative live birth rate (CLBR), and cycle cancelation rate. RESULTS: After propensity score matching, patients with a predicted poor response (antimüllerian hormone, <0.5) on their first in vitro fertilization cycle had similar outcomes on the antagonist protocol (CLBR of 14.2%, 95% confidence intervals [CIs]: 13.6%, 14.8%) compared with flare (CLBR of 13.6%, 95% CIs: 12.4%, 14.8%). We evaluated patients undergoing a second cycle after having a poor response (<4 oocytes retrieved) on their first cycle. Patients in the antagonist-to-antagonist group had a similar change in outcomes between the first and second cycles (average CLBR improvement of 13.9%, 95% CIs: 12.1%, 15.6%) compared with the antagonist-to-flare group (average CLBR improvement of 14.4%, 95% CIs: 10.9%, 18.3%). In addition, patients in the flare-to-antagonist group had a similar change in outcomes between the first and second cycles (average CLBR improvement of 10.4%, 95% CIs: 6.6%, 14.5%) compared with the flare-to-flare group (average CLBR improvement of 9.0%, 95% CIs: 5.1%, 13.4%). CONCLUSION: Poor responders have similar outcomes on an antagonist protocol compared with a flare protocol for both the first and second cycles.

A higher number of oocytes retrieved is associated with an increase in fertilized oocytes, blastocysts, and cumulative live birth rates.

OBJECTIVE: To investigate the association between the number of oocytes retrieved and the numbers of fertilized oocytes and blastocysts and cumulative and primary transfer live birth rates (LBRs). DESIGN: Retrospective study. SETTING: Retrieval cycles and linked embryo transfers from the Society for Assisted Reproductive Technology Clinic Outcome Reporting System. PATIENT(S): Patients in the United States undergoing autologous in vitro fertilization cycles from 2014 to 2019 (n = 402,411 cycles). INTERVENTION(S): None. MAIN OUTCOME MEASURE(S): Normally fertilized oocytes, blastocysts, and cumulative and primary transfer LBRs. RESULT(S): There was a strong positive linear correlation between oocytes and fertilized oocytes and between oocytes and blastocysts. The cumulative LBR increased rapidly with the number of oocytes retrieved to approximately 16-20 oocytes, at which point it continued to increase but with diminishing returns. The increasing trend of the cumulative LBR was observed when stratifying patients by age and antimüllerian hormone and after controlling for confounding variables using multivariate logistic regression. The primary transfer LBR also increased with the number of oocytes to approximately 16-20 oocytes, at which point it plateaued but did not decline. CONCLUSION(S): A higher number of oocytes retrieved improves the cumulative LBR without impairing the primary transfer LBR. This suggests that ovarian stimulation strategies should aim to safely maximize the number of oocytes retrieved.

Wearable Sensors Improve Prediction of Post-Stroke Walking Function Following Inpatient Rehabilitation.

OBJECTIVE: A primary goal of acute stroke rehabilitation is to maximize functional recovery and help patients reintegrate safely in the home and community. However, not all patients have the same potential for recovery, making it difficult to set realistic therapy goals and to anticipate future needs for short- or long-term care. The objective of this study was to test the value of high-resolution data from wireless, wearable motion sensors to predict post-stroke ambulation function following inpatient stroke rehabilitation. METHOD: Supervised machine learning algorithms were trained to classify patients as either household or community ambulators at discharge based on information collected upon admission to the inpatient facility (N=33-35). Inertial measurement unit (IMU) sensor data recorded from the ankles and the pelvis during a brief walking bout at admission (10 meters, or 60 seconds walking) improved the prediction of discharge ambulation ability over a traditional prediction model based on patient demographics, clinical information, and performance on standardized clinical assessments. RESULTS: Models incorporating IMU data were more sensitive to patients who changed ambulation category, improving the recall of community ambulators at discharge from 85% to 89-93%. CONCLUSIONS: This approach demonstrates significant potential for the early prediction of post-rehabilitation walking outcomes in patients with stroke using small amounts of data from three wearable motion sensors. CLINICAL IMPACT: Accurately predicting a patient's functional recovery early in the rehabilitation process would transform our ability to design personalized care strategies in the clinic and beyond. This work contributes to the development of low-cost, clinically-implementable prognostic tools for data-driven stroke treatment.

Validation of Amazon Halo Movement: a smartphone camera-based assessment of movement health.

Movement health is understanding our body's ability to perform movements during activities of daily living such as lifting, reaching, and bending. The benefits of improved movement health have long been recognized and are wide-ranging from improving athletic performance to helping ease of performing simple tasks, but only recently has this concept been put into practice by clinicians and quantitatively studied by researchers. With digital health and movement monitoring becoming more ubiquitous in society, smartphone applications represent a promising avenue for quantifying, monitoring, and improving the movement health of an individual. In this paper, we validate Halo Movement, a movement health assessment which utilizes the front-facing camera of a smartphone and applies computer vision and machine learning algorithms to quantify movement health and its sub-criteria of mobility, stability, and posture through a sequence of five exercises/activities. On a diverse cohort of 150 participants of various ages, body types, and ability levels, we find moderate to strong statistically significant correlations between the Halo Movement assessment overall score, metrics from sensor-based 3D motion capture, and scores from a sequence of 13 standardized functional movement tests. Further, the smartphone assessment is able to differentiate regular healthy individuals from professional movement athletes (e.g., dancers, cheerleaders) and from movement impaired participants, with higher resolution than that of existing functional movement screening tools and thus may be more appropriate than the existing tests for quantifying functional movement in able-bodied individuals. These results support using Halo Movement's overall score as a valid assessment of movement health.

An interpretable machine learning model for individualized gonadotrophin starting dose selection during ovarian stimulation.

RESEARCH QUESTION: Can we develop an interpretable machine learning model that optimizes starting gonadotrophin dose selection in terms of mature oocytes (metaphase II [MII]), fertilized oocytes (2 pronuclear [2PN]) and usable blastocysts? DESIGN: This was a retrospective study of patients undergoing autologous IVF cycles from 2014 to 2020 (n = 18,591) in three assisted reproductive technology centres in the USA. For each patient cycle, an individual dose-response curve was generated from the 100 most similar patients identified using a K-nearest neighbours model. Patients were labelled as dose-responsive if their dose-response curve showed a region that maximized MII oocytes, and flat-responsive otherwise. RESULTS: Analysis of the dose-response curves showed that 30% of cycles were dose-responsive and 64% were flat-responsive. After propensity score matching, patients in the dose-responsive group who received an optimal starting dose of FSH had on average 1.5 more MII oocytes, 1.2 more 2PN embryos and 0.6 more usable blastocysts using 10 IU less of starting FSH and 195 IU less of total FSH compared with patients given non-optimal doses. In the flat-responsive group, patients who received a low starting dose of FSH had on average 0.3 more MII oocytes, 0.3 more 2PN embryos and 0.2 more usable blastocysts using 149 IU less of starting FSH and 1375 IU less of total FSH compared with patients with a high starting dose. CONCLUSIONS: This study demonstrates retrospectively that using a machine learning model for selecting starting FSH can achieve optimal laboratory outcomes while reducing the amount of starting and total FSH used.

An interpretable machine learning model for predicting the optimal day of trigger during ovarian stimulation.

OBJECTIVE: To develop an interpretable machine learning model for optimizing the day of trigger in terms of mature oocytes (MII), fertilized oocytes (2PNs), and usable blastocysts. DESIGN: Retrospective study. SETTING: A group of three assisted reproductive technology centers in the United States. PATIENT(S): Patients undergoing autologous in vitro fertilization cycles from 2014 to 2020 (n = 30,278). INTERVENTION(S): None. MAIN OUTCOME MEASURE(S): Average number of MII oocytes, 2PNs, and usable blastocysts. RESULT(S): A set of interpretable machine learning models were developed using linear regression with follicle counts and estradiol levels. When using the model to make day-by-day predictions of trigger or continuing stimulation, possible early and late triggers were identified in 48.7% and 13.8% of cycles, respectively. After propensity score matching, patients with early triggers had on average 2.3 fewer MII oocytes, 1.8 fewer 2PNs, and 1.0 fewer usable blastocysts compared with matched patients with on-time triggers, and patients with late triggers had on average 2.7 fewer MII oocytes, 2.0 fewer 2PNs, and 0.7 fewer usable blastocysts compared with matched patients with on-time triggers. CONCLUSION(S): This study demonstrates that it is possible to develop an interpretable machine learning model for optimizing the day of trigger. Using our model has the potential to improve outcomes for many in vitro fertilization patients.

A new open-access platform for measuring and sharing mTBI data.

Despite numerous research efforts, the precise mechanisms of concussion have yet to be fully uncovered. Clinical studies on high-risk populations, such as contact sports athletes, have become more common and give insight on the link between impact severity and brain injury risk through the use of wearable sensors and neurological testing. However, as the number of institutions operating these studies grows, there is a growing need for a platform to share these data to facilitate our understanding of concussion mechanisms and aid in the development of suitable diagnostic tools. To that end, this paper puts forth two contributions: (1) a centralized, open-access platform for storing and sharing head impact data, in collaboration with the Federal Interagency Traumatic Brain Injury Research informatics system (FITBIR), and (2) a deep learning impact detection algorithm (MiGNet) to differentiate between true head impacts and false positives for the previously biomechanically validated instrumented mouthguard sensor (MiG2.0), all of which easily interfaces with FITBIR. We report 96% accuracy using MiGNet, based on a neural network model, improving on previous work based on Support Vector Machines achieving 91% accuracy, on an out of sample dataset of high school and collegiate football head impacts. The integrated MiG2.0 and FITBIR system serve as a collaborative research tool to be disseminated across multiple institutions towards creating a standardized dataset for furthering the knowledge of concussion biomechanics.

Rapid Screening of Physiological Changes Associated With COVID-19 Using Soft-Wearables and Structured Activities: A Pilot Study.

OBJECTIVE: Controlling the spread of the COVID-19 pandemic largely depends on scaling up the testing infrastructure for identifying infected individuals. Consumer-grade wearables may present a solution to detect the presence of infections in the population, but the current paradigm requires collecting physiological data continuously and for long periods of time on each individual, which poses limitations in the context of rapid screening. Technology: Here, we propose a novel paradigm based on recording the physiological responses elicited by a short (~2 minutes) sequence of activities (i.e. "snapshot"), to detect symptoms associated with COVID-19. We employed a novel body-conforming soft wearable sensor placed on the suprasternal notch to capture data on physical activity, cardio-respiratory function, and cough sounds. RESULTS: We performed a pilot study in a cohort of individuals (n=14) who tested positive for COVID-19 and detected altered heart rate, respiration rate and heart rate variability, relative to a group of healthy individuals (n=14) with no known exposure. Logistic regression classifiers were trained on individual and combined sets of physiological features (heartbeat and respiration dynamics, walking cadence, and cough frequency spectrum) at discriminating COVID-positive participants from the healthy group. Combining features yielded an AUC of 0.94 (95% CI=[0.92, 0.96]) using a leave-one-subject-out cross validation scheme. Conclusions and Clinical Impact: These results, although preliminary, suggest that a sensor-based snapshot paradigm may be a promising approach for non-invasive and repeatable testing to alert individuals that need further screening.

A Computational Study of Liquid Shock Absorption for Prevention of Traumatic Brain Injury.

Mild traumatic brain injury (mTBI), more colloquially known as concussion, is common in contact sports such as American football, leading to increased scrutiny of head protective gear. Standardized laboratory impact testing, such as the yearly National Football League (NFL) helmet test, is used to rank the protective performance of football helmets, motivating new technologies to improve the safety of helmets relative to existing equipment. In this work, we hypothesized that a helmet which transmits a nearly constant minimum force will result in a reduced risk of mTBI. To evaluate the plausibility of this hypothesis, we first show that the optimal force transmitted to the head, in a reduced order model of the brain, is in fact a constant force profile. To simulate the effects of a constant force within a helmet, we conceptualize a fluid-based shock absorber system for use within a football helmet. We integrate this system within a computational helmet model and simulate its performance on the standard NFL helmet test impact conditions. The simulated helmet is compared with other helmet designs with different technologies. Computer simulations of head impacts with liquid shock absorption predict that, at the highest impact speed (9.3 m/s), the average brain tissue strain is reduced by 27.6% ± 9.3 compared to existing helmet padding when tested on the NFL helmet protocol. This simulation-based study puts forth a target benchmark for the future design of physical manifestations of this technology.

Multi-Directional Dynamic Model for Traumatic Brain Injury Detection.

Given the worldwide adverse impact of traumatic brain injury (TBI) on the human population, its diagnosis and prediction are of utmost importance. Historically, many studies have focused on associating head kinematics to brain injury risk. Recently, there has been a push toward using computationally expensive finite element (FE) models of the brain to create tissue deformation metrics of brain injury. Here, we develop a new brain injury metric, the brain angle metric (BAM), based on the dynamics of a 3 degree-of-freedom lumped parameter brain model. The brain model is built based on the measured natural frequencies of an FE brain model simulated with live human impact data. We show that it can be used to rapidly estimate peak brain strains experienced during head rotational accelerations that cause mild TBI. In our data set, the simplified model correlates with peak principal FE strain (R2 = 0.82). Further, coronal and axial brain model displacement correlated with fiber-oriented peak strain in the corpus callosum (R2 = 0.77). Our proposed injury metric BAM uses the maximum angle predicted by our brain model and is compared against a number of existing rotational and translational kinematic injury metrics on a data set of head kinematics from 27 clinically diagnosed injuries and 887 non-injuries. We found that BAM performed comparably to peak angular acceleration, translational acceleration, and angular velocity in classifying injury and non-injury events. Metrics that separated time traces into their directional components had improved model deviance compare with those that combined components into a single time trace magnitude. Our brain model can be used in future work to rapidly approximate the peak strain resulting from mild to moderate head impacts and to quickly assess brain injury risk.

Vulnerable Locations on the Head to Brain Injury and Implications for Helmet Design.

In studying traumatic brain injury (TBI), it has been long hypothesized that the head is more vulnerable to injury from impacts in certain directions or locations, as the relationship between impact force and the resulting neurological outcome is complex and can vary significantly between individual cases. Many studies have identified head angular acceleration to be the putative cause of brain trauma, but it is not well understood how impact location can affect the resulting head kinematics and tissue strain. Here, we identify the susceptibility of the head to accelerations and brain strain from normal forces at contact points across the surface of the skull and jaw using a three-dimensional, 20-degree-of-freedom rigid-body head and cervical spine model. We find that head angular acceleration and brain tissue strain resulting from an input force can vary by orders of magnitude based on impact location on the skull, with the mandible as the most vulnerable region. Conversely, head linear acceleration is not sensitive to contact location. Using these analyses, we present an optimization scheme to distribute helmet padding thickness to minimize angular acceleration, resulting in a reduction of angular acceleration by an estimated 25% at the most vulnerable contact point compared to uniform thickness padding. This work gives intuition behind the relationship between input force and resulting brain injury risk, and presents a framework for developing and evaluating novel head protection gear.

Passive cervical spine ligaments provide stability during head impacts.

It has been suggested that neck muscle strength and anticipatory cocontraction can decrease head motions during head impacts. Here, we quantify the relative angular impulse contributions of neck soft tissue to head stabilization using an OpenSim musculoskeletal model with Hill-type muscles and rate-dependent ligaments. We simulated sagittal extension and lateral flexion mild experimental head impacts performed on 10 subjects with relaxed or cocontracted muscles, and median American football head impacts. We estimated angular impulses from active muscle, passive muscle and ligaments during head impact acceleration and deceleration phases. During the acceleration phase, active musculature produced resistive angular impulses that were 30% of the impact angular impulse in experimental impacts with cocontracted muscles. This was reduced below 20% in football impacts. During the deceleration phase, active musculature stabilized the head with 50% of the impact angular impulse in experimental impacts with cocontracted muscles. However, passive ligaments provided greater stabilizing angular impulses in football impacts. The redistribution of stabilizing angular impulses results from ligament and muscle dependence on lengthening rate, where ligaments stiffen substantially compared to active muscle at high lengthening rates. Thus, ligaments provide relatively greater deceleration impulses in these impacts, which limit the effectiveness of muscle strengthening or anticipated activations.

Dependency of Head Impact Rotation on Head-Neck Positioning and Soft Tissue Forces.

OBJECTIVE: Humans are susceptible to traumatic brain injuries from rapid head rotations that shear and stretch the brain tissue. Conversely, animals such as woodpeckers intentionally undergo repetitive head impacts without apparent injury. Here, we represent the head as the end effector of a rigid linkage cervical spine model to quantify how head angular accelerations are affected by the linkage positioning (head-neck configuration) and the soft tissue properties (muscles, ligaments, tendons). METHODS: We developed a two-pivot manipulator model of the human cervical spine with passive torque elements to represent soft tissue forces. Passive torque parameters were fit against five human subjects undergoing mild laboratory head impacts with tensed and relaxed neck muscle activations. With this representation, we compared the effects of the linkage configuration dependent end-effector inertial properties and the soft tissue resistive forces on head impact rotation. RESULTS: Small changes in cervical spine positioning (<5 degrees) can drastically affect the resulting rotational head accelerations (>100%) following an impact by altering the effective end-effector inertia. Comparatively, adjusting the soft tissue torque elements from relaxed to tensed muscle activations had a smaller (<30%) effect on maximum rotational head accelerations. Extending our analysis to a woodpecker rigid linkage model, we postulate that woodpeckers experience relatively minimal head impact rotation due to the configuration of their skeletal anatomy. CONCLUSION: Cervical spine positioning dictates the head angular acceleration following an impact, rather than the soft tissue torque elements. SIGNIFICANCE: This analysis quantifies the importance of head positioning prior to impact, and may help us to explain why other species are naturally more resilient to head impacts than humans.

Spinal constraint modulates head instantaneous center of rotation and dictates head angular motion.

The head is kinematically constrained to the torso through the spine and thus, the spine dictates the amount of output head angular motion expected from an input impact. Here, we investigate the spinal kinematic constraint by analyzing the head instantaneous center of rotation (HICOR) with respect to the torso in head/neck sagittal extension and coronal lateral flexion during mild loads applied to 10 subjects. We found the mean HICOR location was near the C5-C6 intervertebral joint in sagittal extension, and T2-T3 intervertebral joint in coronal lateral flexion. Using the impulse-momentum relationship normalized by subject mass and neck length, we developed a non-dimensional analytical ratio between output angular velocity and input linear impulse as a function of HICOR location. The ratio was 0.65 and 0.50 in sagittal extension and coronal lateral flexion respectively, implying 30% greater angular velocities in sagittal extension given an equivalent impulse. Scaling to subject physiology also predicts larger required impulses given greater subject mass and neck length to achieve equivalent angular velocities, which was observed experimentally. Furthermore, the HICOR has greater motion in sagittal extension than coronal lateral flexion, suggesting the head and spine can be represented with a single inverted pendulum in coronal lateral flexion, but requires a more complex representation in sagittal extension. The upper cervical spine has substantial compliance in sagittal extension, and may be responsible for the complex motion and greater extension angular velocities. In analyzing the HICOR, we can gain intuition regarding the neck's role in dictating head motion during external loading.

Head Impact Kinematics Estimation With Network of Inertial Measurement Units.

Wearable sensors embedded with inertial measurement units have become commonplace for the measurement of head impact biomechanics, but individual systems often suffer from a lack of measurement fidelity. While some researchers have focused on developing highly accurate, single sensor systems, we have taken a parallel approach in investigating optimal estimation techniques with multiple noisy sensors. In this work, we present a sensor network methodology that utilizes multiple skin patch sensors arranged on the head and combines their data to obtain a more accurate estimate than any individual sensor in the network. Our methodology visually localizes subject-specific sensor transformations, and based on rigid body assumptions, applies estimation algorithms to obtain a minimum mean squared error estimate. During mild soccer headers, individual skin patch sensors had over 100% error in peak angular velocity magnitude, angular acceleration magnitude, and linear acceleration magnitude. However, when properly networked using our visual localization and estimation methodology, we obtained kinematic estimates with median errors below 20%. While we demonstrate this methodology with skin patch sensors in mild soccer head impacts, the formulation can be generally applied to any dynamic scenario, such as measurement of cadaver head impact dynamics using arbitrarily placed sensors.

Propagation of errors from skull kinematic measurements to finite element tissue responses.

Real-time quantification of head impacts using wearable sensors is an appealing approach to assess concussion risk. Traditionally, sensors were evaluated for accurately measuring peak resultant skull accelerations and velocities. With growing interest in utilizing model-estimated tissue responses for injury prediction, it is important to evaluate sensor accuracy in estimating tissue response as well. Here, we quantify how sensor kinematic measurement errors can propagate into tissue response errors. Using previous instrumented mouthguard validation datasets, we found that skull kinematic measurement errors in both magnitude and direction lead to errors in tissue response magnitude and distribution. For molar design instrumented mouthguards susceptible to mandible disturbances, 150-400% error in skull kinematic measurements resulted in 100% error in regional peak tissue response. With an improved incisor design mitigating mandible disturbances, errors in skull kinematics were reduced to <50%, and several tissue response errors were reduced to <10%. Applying 30[Formula: see text] rotations to reference kinematic signals to emulate sensor transformation errors yielded below 10% error in regional peak tissue response; however, up to 20% error was observed in peak tissue response for individual finite elements. These findings demonstrate that kinematic resultant errors result in regional peak tissue response errors, while kinematic directionality errors result in tissue response distribution errors. This highlights the need to account for both kinematic magnitude and direction errors and accurately determine transformations between sensors and the skull.