59^th IBSIS-RMBS
the joint
59^th International Biomedical Sciences Instrumentation Symposium
&
59^th Rocky Mountain Bioengineering Symposium

The loading experienced by military helicopter and fighter aircrew can lead to spinal injury by degrading the soft tissues. Fighter pilots endure low frequency high magnitude loading, while helicopter pilots endure high frequency vibrational loading over longer periods of time. The purpose of this study is to analyze intervertebral disc material properties, including elastic modulus, failure strain, and failure stress. This will provide data to model the annulus fibrosus to predict the risk of injuries. Quantifying the risk to specific spinal locations could provide insight to improve the design of protective clothing and equipment. Single layers of the porcine annulus fibrosus were subjected to tensile failure tests to determine the mechanical properties. Results showed a significantly (p<0.05) higher elastic modulus in the circumferential direction compared to the axial direction. There was also a significantly greater elastic modulus for samples from the lateral region of the disc compared to the posterior region. There was no significant difference for the anterior region or specimen depth with regards to elastic modulus. Failure stress showed significant differences between superficial and deep specimens as circumferential and axial testing directions. Failure strain only had differences between circumferential and axial directions. This data can be used to help predict where in the disc an injury may occur. Because of the decreased posterior elastic modulus, tensile failure is more likely to occur in that region.

Breast cancer is the most commonly diagnosed disease in women as well as the leading cause of cancer death. Neoadjuvant chemotherapy (NAC) is a standard treatment technique for locally advanced breast tumors, with the goal of clinical downstaging. Dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) is used to diagnose breast cancer and could be used to access the NAC treatment response. However, the assessment of treatment response at the early phase is crucial and challenging. Previous studies show that radiomic features have the potential in quantifying the pathological changes due to NAC. In this work, an attempt has been made to objectively quantify the early phase changes observed on the DCE-MRI due to NAC using radiomic features. Different radiomic features are extracted from DCE-MR images of Visit 1 & 2 to quantify the treatment response and are statistically analyzed. Further, kinetic features from the images are also extracted and associated with the radiomic features and are tested using Spearman's correlation test. Results demonstrate that radiomic features such as energy, homogeneity, entropy, kurtosis have shown a highly significant difference between the two visits (p = 0.05). Similarly, the association between significant radiomic features and kinetic parameters shown a positive correlation with r = (0.48 - 0.66), p = (0.00029 - 0.016) respectively. Hence it appears that the association of radiomic and kinetic parameters would be used as an adjunct measure in differentiating the NAC treatment visits at early stages.

Prediction techniques are extensively used in medical applications and health care devices. The prediction of the infusion flow rate for the required drug dosage and drug concentration in a smart wireless infusion pump is necessary for precise drug flow for the patients. In this paper, the prediction model has been developed to predict the lag time using Gaussian Process Regression (GPR) technique with a squared exponential kernel. Currently, a smart wireless infusion pump is incorporated with its smart drug library. The required parameters such as drug dosage, drug flow rate are utilized as inputs to predict the lag time and to minimize start-up delays using the proposed regression technique. The evaluation of the prediction model is done by the coefficient of determination (R2), mean absolute error (MAE), and root-mean-squared error (RMSE). These prediction results are verified for predicting lag time for two different carrier flowrates 10 ml/hr and 50 ml/hr. The outcome of the study indicates that the regression model GPR has better prediction accuracy with a mean R2 of 0.9. Hence, the GPR technique is capable to achieve quick infusion and optimal flow rate with minimized lag time for smart infusion pumps.

Deep vein thrombosis (DVT) refers to the formation of blood clots in the deep venous system of the body. Although DVT can occur in any vein in the body, it is commonly seen in the veins of the lower extremity and pelvis. Stasis of blood flow, endothelial injury and hypercoagulable states are factors that significantly accelerate the formation of clots in the veins. Pregnancy and puerperium are physiologic hypercoagulable states that predisposes women to the DVT and its complication during the normal childbearing process. The spectrum of DVT varies from mild, asymptomatic to severe life-threatening consequences like pulmonary embolism (PE), resulting from a dislodged clot in the pulmonary circulation leading to respiratory compromise and death. Thus, early, and accurate diagnosis with appropriate management is crucial in preventing mortality from DVT-PE. Currently, clinical suspicion, blood tests and radiological evidence are used to diagnose DVT/PE based on the patient factors. Each of the modalities have several shortcomings of limited feasibility during gestation to reduced diagnostic sensitivity. The novel technological advancement of this device utilizes near-infrared radiation to emit light onto an impacted limb and read the absorption. Oxygenated hemoglobin [HbO2] absorbs NIR light, and a clot will result in reduced [HbO2] flowing through the impacted limb. Consequently, resulting in reduced absorption of NIR when compared to a healthy limb giving objective diagnostic prior to physiological symptoms. Testing of the instrumentations allowed for signal detection of 3 subjects. Further large-scale testing will allow for validation of NIR as a diagnostic tool for DVT.

The Corpus Callosum (CC) is a large white matter bundle that connects the left and right cerebral hemispheres of the human brain. It is susceptible to atrophy as Alzheimer's disease progresses. The robust segmentation of CC allows quantitative investigation of its structural changes. However, deep learning-based CC segmentation is less explored. In this work, an improved UNet model is proposed for CC segmentation from two-dimensional T1-weighted mid-sagittal brain MRI. For this, mid-sagittal scans (n = 184) from the publicly available Open Access Series of Imaging Studies (OASIS) brain MRI database are used. The images are fed to an improved UNet++ network. The architecture contains a fully convolutional network with two paths, contracting and extracting, that are connected in a U-shape to automatically extract spatial information. Leave one out Cross Validation (LooCV) method is used to evaluate the robustness of the proposed method. Results show that the proposed approach is able to segment CC from MR images. The proposed method yields the Dice score of 98.43%, and Jaccard index of 98.53%. The improved UNet++ model obtained the highest sensitivity of 99.21% for AD conditions. Further, the performance of the proposed model has been validated against the state-of-the-art methods. Thus, the proposed approach could be useful for the segmentation of MR images in clinical condition.

Cardiovascular diseases (CVD) indicate various condition that causes irregular heart rhythm (arrhythmia) and leads to a heart attack. An estimated 18 million people die from CVDs every year, which is 32% of deaths worldwide. The Electrocardiogram (ECG) is an essential tool that aids in detecting arrhythmia. As visual inspection of ECG is challenging due its intricate details, a computer aided system is necessary to distinguish normal and arrhythmic signals. Deep learning (DL) provides an efficient way to detect the minute details in an ECG. In this work, DL models based on residual neural network (ResNet), with 50 and 34 layers, have been built and optimized for classification. The model utilizes PTB-XL ECG dataset containing 21837 clinical 12-lead signals. The ResNet50 has 48 convolutional layers along with an Average Pooling and a MaxPooling layer while ResNet34 has 32 convolutional layers instead of 48. The fully connected network contains 3 layers with neurons 8, 2, and 1. The training is performed using Adam optimizer with a learning rate of 0.0005 for 50 epochs, and the performance metrics are obtained. The trained ResNet50 and ResNet34 can distinguish the ECG records into arrhythmia and non-arrhythmia with an accuracy of 86.24% and 81.60% respectively. The percentage difference in F1 score of ResNet50 with respect to ResNet34 is 6.46%. The increased number of layers led to better feature extraction and performance. The ResNet50 model could effectively classify the arrhythmia and non-arrhythmia records compared to ResNet34 and can be exploited to identify the source of arrhythmia.

Understanding neuronal structure and function is essential to studying the human brain. The Air Force Research Laboratory (AFRL) has funded this project to create a model of human brain neurons that accurately reflects neuronal function, energy consumption, and oxygen consumption. Extensive work has been performed on the Hodgkin-Huxley model of neurons to accurately model neuronal firing. This study focuses on the creation of a model of the Hodgkin-Huxley neuron in MATLAB with the assistance of the DynaSim toolbox. This model was used to compute values and using another model the energy efficiency of the neuron was calculated and related to oxygen consumption, which then corresponds to Blood Oxygenation Level Dependent (BOLD) imaging in fMRI. The results of this study provide detailed visual and technical information about how the brain neurons function, which is crucial to the further development in brain imaging techniques and to further our understanding of why and how our complex brains function.

Estrogen Receptor (ER) is a molecular biomarker that plays an important role in evaluating the Neoadjuvant Chemotherapy (NAC) treatment response of breast cancer patients. ER (-) breast cancer patients have better tumor response rates than ER (+) patients due to NAC and the result of ER status could change after NAC. However, there are limited studies on the analysis of NAC treatment response using ER status. Further, manual quantification of treatment response is challenging and inconsistent across raters. In this work, an attempt has been made to objectively quantify the radiological differences of Dynamic Contrast Enhanced (DCE) MR images in ER (-) and ER (+) patients due to NAC using Gabor filter derived Anisotropy Index (AI). The images (113 subjects at 4 visits of NAC treatment) used in this study are obtained from the publicly available I-SPY1 dataset. Gabor filter bank is designed with 5 scales and 7 orientations, and AI is calculated from each Gabor energy within the patient group. Results show that AI values can statistically (p<0.05) differentiate the radiological differences in ER (-) and ER (+) patients due to NAC. The percentage difference in the mean AI values of Visit 1 Vs Visit 4, Visit 1 Vs Visit 3, and Visit 2 Vs Visit 4 is high in ER (-) compared to ER (+) patients. Thus, Gabor filter derived AI could be used as an objective measure in evaluating NAC treatment response in ER (-) and ER (+) patients.

Breast cancer is the most predominant disease and foremost cause of cancer deaths in women worldwide, with treatment plans varying regardless of the grade and biology of the tumor. Neoadjuvant chemotherapy (NAC) is the standard clinical implementation to retrench the tumor size and escalate the breast-conserving rate. Dynamic contrast enhanced MR imaging (DCE-MRI) is an effective modality in analyzing the response during NAC treatment. However lower-grade cancer patients are slow growing tumors with a better prognosis, but higher-grade cancer patients aggressively grow and require effective treatment. So, it is necessary to investigate the grade specific response information during the NAC treatment. In this work, analysis of NAC treatment response on breast cancer patients is performed by investigating the low- and high-grade cancer patients separately using DCE MR images. Twenty-six patient data with three visits of NAC treatment is obtained from QIN BREAST and QIN BREAST-02 datasets from the openly available TCIA database. The mean intensity (MI) value is calculated from manually segmented tumor volumes at different visits for both low- and high-grade cancer patients. The results demonstrate that mean intensity values showed a statistical difference between Visit 1 & 3 in both low- and high-grade patients during NAC with p = 0.05. The percentage difference in mean intensity value between Visit 1 & 3 of high-grade subjects is observed to be high compared to low-grade subjects. Hence it appears that the high-grade breast cancer patients respond well to NAC treatment response compared to low-grade breast cancer patients.

Alzheimer's Disease (AD) is a progressive irreversible neurodegenerative disorder which involves the deformations in brain sub-anatomic regions. Recent studies suggest that these deformations could be characterized using bi-planar information extracted from structural Magnetic Resonance (MR) image features. However, analysis and fusion of these bi-planar features have been a challenging task in AD differentiation. In this study, an attempt has been made to fuse the characteristics of axial and sagittal view MR images using Canonical Correlation Analysis (CCA) for the differentiation of Healthy Controls (HC) and AD. For this, MR brain images obtained from a public database are skull stripped and spatially registered. Morphometric features are extracted from the pre-processed mid-sagittal and mid-axial images using histogram of oriented gradients. Further, these extracted features are fused using CCA. The performance of classifier is analyzed for the variations in canonical component dimensions. Results indicate that the morphometric feature spaces extracted from sagittal and axial planes individually overlap for HC and AD. The proposed CCA based fusion of sagittal and axial features exhibit variations between HC and AD images for a canonical feature dimension of 30. Performance of the adopted approach confirms that the bi-planar feature fusion is essential for the differentiation of AD.

In this work, an attempt has been made to classify dichotomous emotional states using EDA and geometric features. For this, the annotated happy and sad EDA are obtained from the online public database. The EDA is subjected to discrete Fourier transform and Fourier coefficients in complex plane are obtained. The envelop of the complex plane is identified using a-shape method. Five geometric features, namely centre of gravity, eccentricity, convexity, rectangularity, and convex hull area are computed from the envelop and statistical analysis is performed. Two machine-learning algorithms, namely support vector machine and random forest (RF) are considered for the classification. The results show that the proposed approach is able classify the dichotomous emotional states. The rectangularity feature is found to be distinct and show statistically significant difference between the happy and sad emotional states (p<0.05). The RF classifier yields a highest F-m and AUC of 87.8% and 93.8%, respectively in differentiating emotional states. Thus, it appears that the proposed method could be used to understand the neurological, psychiatrical, and bio behavioural mechanisms associated with happy and sad emotional states.

The male version of lacrosse is a helmeted collision sport, which allows stick checking and body contact. Although changes to checking rules have attempted to reduce the incidence of concussion in male lacrosse, head impacts can occur when stick checking and body contacts are being executed. Few studies have used rigorous video analysis to verify sensor recorded events in male lacrosse. A male high school varsity lacrosse team of 25 players wore the Stanford Instrumented Mouthguard (MiG) during 12 competitive games. Video footage was reviewed to remove false-positive recordings and verify head impacts, which resulted in 13 head impacts over 69 athlete-exposures (0.19 head impacts per athlete-exposure). Of the 13 video-confirmed head impacts, 7 impact events were stick-to-head (53.8%), 3 were body-to-head (23.1%), 1 was head-to-head (7.7%) and the remaining 2 events were indirect impacts involving body-to-body contact with no observable head contact (15.4%). The most common impact site was the front of the head (5, 38.5%), followed by the side (3, 23.1%) and rear (3, 23.1%). The median peak linear acceleration, angular velocity and angular acceleration values of the 13 video-verified impacts were 36.1 g, 13.7 rad/s and 3596 rad/s2, respectively. Future work should be directed towards the collection of a larger sample size of impacts in male lacrosse to confirm the estimates of head impact biomechanics reported in the current study. In addition, substitutions should be temporally tracked so that impact rate can be calculated per player-hour for more accurate comparisons across sports and gender. Testing of the instrumentations allowed for signal detection of 3 subjects. Further large-scale testing will allow for validation of NIR as a diagnostic tool for DVT.

Injuries from BB shots are responsible for thousands of injuries each year, with many resulting in contusions or superficial embedments to the extremities. To help better understand the injury biomechanics of BB shots, a porcine foreleg finite element model was selected for comparison with some documented porcine ballistic experiments. The model was created by segmenting a porcine leg computerized tomography (CT) scan into the major bones, skin, and soft tissues, and then generating a mesh from the resulting geometries. A previously published hyperelastic material model was incorporated to represent the skin’s non-linear mechanical behavior. The pig leg model was used to simulate the skin response to 87.1 and 114.6 m/s stainless steel BB shots. The simulation matched the non-penetrative behavior from the experiments, predicting peak dynamic deformations of 12.6 and 25.4 mm respectively. MatLab was used to collect the movement of surface nodes and reconstruct continuous surfaces every 0.5 ms. The position and speed of the impact-induced wave was non-linear and did not depend on the BB initial velocity. Future work is needed to compare the simulation results against experimental digital image correlation (DIC) data, increase the time and spatial resolution of simulated sampling surface, and eventually include dynamic material data to account for skin damage with increasing BB initial velocity.

The aim of this study was to provide proof of feasibility in using in-ear accelerometers to monitor head motion during a blast event. Blast tests were performed with a shock tube and an isolated Hybrid III (H3) head/neck combination that was subject to a frontal blast in each trial. An accelerometer package was placed inside the H3 head at the center of gravity (CG) and sagittal plane accelerations along the anterior-posterior and superior-inferior axis of motion were recorded. Two uniaxial accelerometers were placed inside of a foam ear plug and then placed inside of a simulated ear canal on the H3 head and also measured accelerations along the same axes. The results of the blast test trials displayed that absolute peak in-ear accelerations and head CG accelerations correlated well, with r2 values of 0.99 and 0.96 for the two axes of motion in consideration. Although this study does not translate in-ear accelerations to the head CG via rigid body dynamics or automatically filter the high-frequency content of the shock front, the feasibility of using in-ear accelerometers for blast situations is still well demonstrated. This study further provides context to the necessary software required for signal processing of shock fronts.

Injuries from BB shots are responsible for thousands of injuries each year, with many resulting in contusions or superficial embedments to the extremities. To help better understand the injury biomechanics of BB shots, a porcine foreleg finite element model was selected for comparison with some documented porcine ballistic experiments. The model was created by segmenting a porcine leg computerized tomography (CT) scan into the major bones, skin, and soft tissues, and then generating a mesh from the resulting geometries. A previously published hyperelastic material model was incorporated to represent the skin’s non-linear mechanical behavior. The pig leg model was used to simulate the skin response to 87.1 and 114.6 m/s stainless steel BB shots. The simulation matched the non-penetrative behavior from the experiments, predicting peak dynamic deformations of 12.6 and 25.4 mm respectively. MatLab was used to collect the movement of surface nodes and reconstruct continuous surfaces every 0.5 ms. The position and speed of the impact-induced wave was non-linear and did not depend on the BB initial velocity. Future work is needed to compare the simulation results against experimental digital image correlation (DIC) data, increase the time and spatial resolution of simulated sampling surface, and eventually include dynamic material data to account for skin damage with increasing BB initial velocity.

Long duration spaceflight missions can affect the cognitive and behavioral activities of astronauts due to changes in gravity. The microgravity significantly impacts the central nervous system physiology which causes the degradation in the performance and lead to potential risk in the space exploration. The aim of this study was to evaluate functional connectivity at simulated space conditions using an unloading harness system to mimic the body-weight distribution related to Earth, Mars, and International Space Station. A unity model with six directional arrows to imagine six different motor imagery tasks associated with arms and legs were designed for the Oculus Rift S virtual reality headset for testing. An Electroencephalogram (EEG) and functional near infrared spectroscopy (fNIRS) signals were recorded from 10 participants in the distributed weight conditions related to Earth, Mars, and International Space station using the g.Nautilus fNIRS system at sampling rate of 500 Hz. The magnitude squared coherence were estimated from left vs right hemisphere of the brain that represents functional connectivity. The EEG coherence was the higher which shows the strong functional connectivity and fNIRS coherence was lower shows weak functional connectivity between left vs right hemisphere of the brain, during all the tasks and trials irrespective of the simulated space conditions. Further analysis of functional connectivity needed between the inter regions of the brain.

Brain activity in pain-related regions of the cortex region of the brain is thought to be influenced by interactions between expectations and incoming sensory information. However, the subjective experience of pain based on expectation has not been thoroughly explored. By using changes in visual and auditory cues to set expectation levels, a relationship between unconscious expectation of pain and actual sensory pain experienced was characterized. With IRB approval, forty-one volunteers were randomly and blindly grouped into low and high pain expectancy groups. Each group completed a survey about initial pain expectation before the test followed by a survey about actual pain perception after the test. Each group was given verbal instructions from a high or low pain expectancy script, respectively. Then the subjects submerged an arm in an ice bath. The duration of pain threshold (time to feel pain) and the pain tolerance (time to remove hand from ice water) were measured in seconds. The maximum time of exposure was limited to 900 seconds for safety reasons. The results reveal that individuals with higher unconscious expectation levels of pain experienced significant increases (p<0.05) in how quickly the feeling of pain began and the severity of perceived pain. These results confirm that an expectation of impending pain can significantly alter the formation of actual pain experienced and provide insight into how positive expectations can lead to decreased pain felt in chronic disorders and diseases.

Oculomotor measures are utilized for various purposes, including the diagnoses of health conditions. However, less is known about eye movements and their relationship to postural stability, and no consensus is found in the literature on this subject. This retrospective chart review of refractory neurological cases focused on the relationship between postural control as measured by posturography and the characteristics of saccade and pursuit movements. Results included: the Head Neutral average stability score being higher than Head Left or Right; smooth pursuit gain being lower at 0.1 Hz than at 0.4 Hz; in the saccadic tests, the considered subjects tended to undershoot, and the latency tended to be lower than the normative 200 ms; there was a statistically significant difference of the smooth pursuit results with the head neutral and turned right, with very poor performance of the postural control system associated with very poor oculomotor ability to follow a target; only five out of 144 correlation pairs were found statistically significant, pointing toward no correlation between oculomotor activities and postural stability. Although results were conflicting, this study highlighted the need for better testing protocol: what is feasible with healthy subjects in a laboratory setting is not always applicable in clinical settings with subjects affected by different pathologies and in the presence of multiple comorbidities. Furthermore, since the oculomotor system is one component of the postural control system, a greater neurophysiological understanding of how it integrates the inputs received from its subsystems may be needed to lead to more effective interventions.

Collegiate level baseball players routinely pitch from behind an L-screen in a batting cage during indoor batting practice. The purpose of the L-screen is to serve as a guard protecting the pitcher from being struck by a batted ball. However, the screen provides a false sense of security that the pitcher will be able to duck behind the screen upon a line drive hit. Human factors is the study of how humans interact with their environment. This includes studying factors such as reaction times needed for pitchers to safely react to a batted ball. The greater the number of choices, the larger the body segment that must be moved, and the distance that must be moved are all factors that increase the amount of time required to react to a batted ball. Thus, moving the torso and head behind the L-screen after the ball is hit requires significantly more time than moving only the hand. Moreover, the lower arousal and vigilance of a pitcher who feels safe behind a guard performing a repetitive task in practice also leads to slower reaction times. Therefore, while live-pitching in the batting cage, a college baseball pitcher does not have sufficient reaction time to move behind the screen after the ball is hit to avoid being struck. Therefore, training to stay behind the L-screen while pitching is required to eliminate the need to rely on high levels of attentiveness and reaction time during the repetitive task of pitching during batting practice to avoid injuries.

Preterm (gestational age < 37 weeks) birth has significant emotional and financial impacts for both families and society. Accurate detection will allow physicians to begin early treatments and avoid complications such as neonatal developmental disorders, hospitalization and inappropriate treatment. In this study, an attempt has been made to detect the preterm condition using electrohysterography (EHG) signals and strip spectral correlation algorithm (SSCA). For this, the signals are obtained from a public database. SSCA is applied to calculate spectral correlation density (SCD). Statistical features such as, mean, variance, skewness, and kurtosis are derived from the SCD and statistically analyzed. The result show that the extracted SCD spectrum reveals the presence of cyclostationarity in these signals. It is also observed that three features namely, mean, variance, and kurtosis show significant difference between term and preterm conditions. Among these features, kurtosis have less variability in detecting preterm birth. Hence, it appears that cyclostationary based statistical features can be useful in detecting preterm birth.