Indiana University-Purdue University Indianapolis (IUPUI) / Biomechanical Big Data is of great significance to precision health applications, among which we take special interest in Physical Activity Detection (PAD). In this study, we have performed extensive research on deep learning-based PAD from biomechanical big data, focusing on the challenges raised by the need for real-time edge inference. First, considering there are many places we can place the motion sensors, we have thoroughly compared and analyzed the location difference in terms of deep learning-based PAD performance. We have further compared the difference among six sensor channels (3-axis accelerometer and 3-axis gyroscope). Second, we have selected the optimal sensor and the optimal sensor channel, which can not only provide sensor usage suggestions but also enable ultra-lowpower application on the edge. Third, we have investigated innovative methods to minimize the training effort of the deep learning model, leveraging the transfer learning strategy. More specifically, we propose to pre-train a transferable deep learning model using the data from other subjects and then fine-tune the model using limited data from the target-user. In such a way, we have found that, for single-channel case, the transfer learning can effectively increase the deep model performance even when the fine-tuning effort is very small. This research, demonstrated by comprehensive experimental evaluation, has shown the potential of ultra-low-power PAD with minimized sensor stream, and minimized training effort. / 2023-06-01

The high prevalence of work-related musculoskeletal disorders in occupational settings necessitates the development of economic, accurate, and convenient methods for quantifying biomechanical risk exposures. In terms of lifting, the occupational work environment does not provide resources for recording the start and end times of lifting tasks performed by individual workers. As a result, automatic detection of lift starts and ends is required for practical purposes. Occupational lifting styles vary depending on the asymmetry angle, which is the degree of shoulder or trunk rotation required by the lifting task. Predictive or machine learning (ML) algorithms have been increasingly used in the ergonomics field to identify occupational risk factors, such as lifting loads. However, such algorithms are often developed and validated using the dataset collected from the same lab-based experimental set-up, which limits their external validity. The recent development of wearable armbands with surface electromyography (sEMG) electrodes provides a low-cost, wireless, and non-invasive way to collect EMG data beyond laboratory settings. Despite their tremendous potential for field-based workload estimation, these armbands have not been widely implemented yet in automated lift detection and occupational workload estimation. The objective of this study was to evaluate the performance of machine learning (ML) algorithms in the automatic detection of lifts and classification of hand loads during manual lifting tasks from the data acquired by a wearable armband sensor with eight surface electromyography (sEMG) electrodes. Twelve healthy participants (six male and six female) performed repetitive symmetric (S), asymmetric (A), and free dynamic (F) lifts with three different hand-load levels (5 lb, 10 lb and 15 lb) at two origin (24" and 36") and two destination heights (6" and 36"). Three ML algorithms were utilized: Random Forest (RF), Support Vector Machines (SVM) and Gaussian Naïve Bayes (GNB). For lift detection, a subset of four participants was analyzed as a preliminary investigation. RF showed the best performance with the mean start and end errors of 0.53 ± 0.25 seconds and 0.76 ± 0.28 seconds, respectively. The accuracy score of 84.3 ± 3.3% was reported for lift start and 83.3 ± 9.9% for lift end. For hand-load classification, prediction models were developed using four different lifting datasets (S, A, S+A, and F) and were cross-validated using F as the test dataset. Mean classification accuracy was significantly lower in models developed with the S dataset (78.8 ± 7.3%) compared to A (83.3 ± 7.2%), S+A (82.1 ± 7.3%), and F (83.4 ± 8.1%). Overall, findings indicate that the implementation of ML algorithms with wearable EMG armbands for automatic lift detection in occupational settings can be promising. In hand-load classification, models developed with only controlled symmetric lifts were less accurate in predicting loads of more dynamic, unconstrained lifts, which is common in real-world settings. However, since both A and S+A demonstrated equivalent model accuracy with F, EMG armbands possess strong potential for estimating the hand loads of free-dynamic lifts using constrained lift trials involving asymmetric lifts. / Master of Science / Naturalistic occupational settings involve prolonged, frequent, and physically heavy lifting-lowering tasks that are associated with a high prevalence of musculoskeletal disorders. This necessitates the development of economic, accurate, and convenient methods for quantifying risk exposures such as load magnitude, repetitiveness and duration. In terms of lifting, the occupational work environment does not provide resources for recording the start and end of lifting tasks performed by individual workers for analysis. As a result, automatic detection of lift starts and ends is required for practical purposes. Occupational lifting styles vary depending on the asymmetry angle, which is the degree of shoulder or trunk rotation required by the lifting task. Predictive or machine learning (ML) algorithms have been increasingly used in the ergonomics field to identify occupational risk factors, such as lifting loads. However, such algorithms are often developed and validated using the dataset collected from the same lab-based experimental set-up, which limits their external validity. Electromyographic (EMG) signals representing the neuromuscular activity associated with muscular contractions can be valuable for exposure assessment. The recent development of wearable armbands with surface electromyography (sEMG) electrodes provides a low-cost, wireless, and non-invasive way to collect EMG data beyond laboratory settings. Despite their tremendous potential for field-based workload estimation, these armbands have not been widely implemented yet in automated lift detection and occupational workload estimation. The objective of this study was to evaluate the performance of machine learning (ML) algorithms in the automatic detection of lifts and classification of hand loads during manual lifting tasks from the data acquired by a wearable armband sensor with eight surface electromyography (sEMG) electrodes. Twelve healthy participants (six male and six female) performed repetitive symmetric (S), asymmetric (A), and free dynamic (F) lifts with three different hand-load levels (5 lb, 10 lb and 15 lb) at two origin (24" and 36") and two destination heights (6" and 36"). Three ML algorithms were utilized: Random Forest (RF), Support Vector Machines (SVM) and Gaussian Naïve Bayes (GNB). For lift detection, a subset of four participants was analyzed as a preliminary investigation. RF showed the best performance with the mean start and end errors of 0.53 ± 0.25 seconds and 0.76 ± 0.28 seconds, respectively. The accuracy score of 84.3 ± 3.3% was reported for lift start and 83.3 ± 9.9% for lift end. For hand-load classification, prediction models were developed using four different lifting datasets (S, A, S+A, and F) and were cross-validated using F as the test dataset. Mean classification accuracy was significantly lower in models developed with the S dataset (78.8 ± 7.3%) compared to A (83.3 ± 7.2%), S+A (82.1 ± 7.3%), and F (83.4 ± 8.1%). Overall, findings indicate that the implementation of ML algorithms with wearable EMG armbands for automatic lift detection in occupational settings can be promising. In hand-load classification, models developed with only controlled symmetric lifts were less accurate in predicting loads of more dynamic, unconstrained lifts, which is common in real-world settings. However, since both A and S+A demonstrated equivalent model accuracy with F, EMG armbands possess strong potential for estimating the hand loads of free-dynamic lifts using constrained lift trials involving asymmetric lifts.

Acute congestive heart failure is a leading cause of morbidity and mortality. Patients often undergo repeated hospitalizations with an annual cost in excess of $32B dollars. Early detection of impending acute congestion allows for pharmaceutical interdiction that prevents hospitalization, improves outcomes, and reduces healthcare spending. A subcutaneous implantable monitoring device that detects impending acute congestive heart failure by using heart sounds may provide a valuable tool that can be used to titrate heart failure medications to prevent acute heart failure requiring hospitalization. The device may be used to measure changes in the intervals between the R-wave and S1 and S2 heart sounds, and to detect evolving S3 and S4 heart sounds consistent with volume overload. The amplitudes of S1 and S3 heart sounds change as heart failure patients' symptoms worsen. Designing a non-invasive, external device, capable of recording heart sounds and EKGs in patients undergoing pharmaceutical regression of acute congestive heart failure in a hospital setting may give important insight into the nature of heart sound and EKG changes that occur in patients during progression of acute heart failure while they lead their day-to-day lives. This thesis details the design of a portable, non-invasive device, worn externally on the left-pectoral muscle, capable of continuously recording human EKG signals and heart sounds (through custom MEMS accelerometer technology) over a period of two days. Hardware testing for the scope of this thesis exclusively involved healthy volunteers.

The construction industry, one of the largest employers of labor in the United States, has long suffered from health and safety issues relating to work-related musculoskeletal disorders. Back-related injuries are one of the most prevalent of all musculoskeletal disorders in the construction industry. Due to advancements in the field of wearable technologies, wearable robots such as passive back-support exoskeletons have emerged as a possible solution. Exoskeletons have the potential to augment human capacity, support non-neutral work positions, and reduce muscle fatigue and physical exertion. Current research efforts to evaluate the potential of exoskeletons in other industry sectors have been focused on outcome measures such as muscle activity, productivity, perceived discomfort and exertion, usability, and stakeholders' perspectives. However, there is scarce evidence regarding the efficacy of using exoskeletons for construction work. Furthermore, the risks and sociotechnical challenges of employing exoskeletons on construction sites are not well documented. Thus, through the lens of human-centric and socio-technical considerations, this study explores the prospects of adopting back-support exoskeletons in the construction industry. Firstly, a laboratory experiment was conducted to quantify the impact of using a passive exoskeleton for construction work in terms of muscle activity, perceived discomfort, and productivity. In order to investigate the acceptance of exoskeletons among construction workers and the challenges of adopting exoskeletons on construction sites, field explorations evaluating usability, perceived discomfort and exertion, social influence, and workers user perceptions were executed. Using sequential mixed methods approach, the stakeholders and factors (i.e., facilitators and barriers) critical for the adoption of exoskeletons on construction sites were investigated. Thereafter, by employing the factors and leveraging the constructs of the normalization process theory, an implementation plan to facilitate the adoption of passive exoskeletons was developed. The study contributes to the scarce body of knowledge regarding the extent to which exoskeletons can reduce ergonomic exposures associated with construction work. This study provides evidence of the perceptions of the contextual use of wearable robots, and workers' interaction with wearable robots on construction sites. The study contributes to the normalization process theory by showing its efficacy for the development and evaluation of implementation frameworks for construction industry. Furthermore, this study advances the socio-technical systems theory by incorporating all its subsystems (i.e., human, technology, organization and social) for investigating the potential of using a passive back support exoskeleton in the construction industry. / Doctor of Philosophy / Construction workers are often subjected to harsh working conditions and physically demanding work postures, which are ergonomics risks causing back-related musculoskeletal injuries. These injuries have the potential to cause permanent disabilities, lead to early retirement of experienced labor, and is one of the causes of the shortage of skilled workforce in construction. Wearable robots, such as passive back-support exoskeletons, are increasingly been looked upon as a potential solution to mitigate the problem. Exoskeletons are wearable technologies that can support and reinforce workers' body parts. Studies have shown that the use of exoskeletons could lead to reduced muscle fatigue thereby decreasing injuries in the long run. However, most of the research on the use of exoskeletons is focused on other industrial sectors. Scarce evidence regarding the use of exoskeletons in construction is documented in the literature. Furthermore, the use of exoskeletons on construction sites could have certain unintended consequences. Thus, the objective of this research was to understand the risks and challenges of using passive exoskeletons in the construction industry. A laboratory experiment was conducted to measure the impact of using exoskeletons on physical demand and productivity while performing construction tasks. An increase in productivity and a reduction in discomfort in the lower back were observed while using an exoskeleton. Thereafter, field studies were conducted where construction workers performed their usual tasks using an exoskeleton to understand their user experience and acceptance. To help construction companies in the adoption of exoskeletons, facilitators and barriers to the adoption of exoskeletons were identified. Thereafter a plan was developed to facilitate the implementation of passive exoskeletons in construction organizations. This plan can guide construction companies in the adoption of passive exoskeletons. The outcomes of this study will help other researchers to conduct similar studies with other wearable technologies.

The construction industry is one of the most unsafe industries for workers in the United States. Advancements in wearable technology have been proven to create a safer construction environment. Despite the availability of these devices, use within the construction industry remains low. The objective of this research is to identify and analyze the causes behind the reluctance of the construction industry to implement two specific wearable safety devices, a biometric sensor, and a location tracking system. Device acceptance was analyzed from the perspective of the user (construction field labor) and company decision makers (construction managers). A modified unified theory of acceptance and use of technology (UTAUT) model was developed specific to barriers commonly found within technology adoption in the construction industry including: perceived performance expectancy, perceived effort expectancy, openness to data utilization, social influence, data security, and facilitating conditions. A structured questionnaire was designed to test for association between the mentioned constructs and either behavioral intention or actual use. The questionnaire went through an expert review process, and a pilot study was conducted prior to being distributed to industry. Once all data was received Pearson chi-squared analysis was used to test for association between the constructs. A minority (46%) of labor respondents would not agree to voluntarily use the biometric wearable sensing device. Constructs associated with this finding included perceived performance expectancy, perceived effort expectancy, and social influence. A majority (59%) of labor respondents would not agree to voluntarily use the location tracking wearable sensing device. Constructs associated with this finding included perceived performance expectancy, social influence, and data security. A majority (56%) of management respondents would not implement the biometric wearable sensing device. Constructs found to be associated with this finding included perceived performance expectancy, openness to data utilization, and social influence of the client. A supermajority (68%) of management respondents would not implement the location tracking wearable sensing device. Constructs found to be associated with this finding include perceived performance expectancy, perceived effort expectancy, openness to data utilization, social influence, and data security. This study will aid in the successful implementation of wearable sensing devices within the construction industry. Findings from this study can be used to aid those hoping to implement wearable sensing devices by identifying causes of wearable sensing device rejection. The results of this study can be used by both project managers and health and safety professionals to aid in device acceptance by field labor, and by those whose goal is to increase device use among construction firms.

Back pain and injury are a global health issue and are a leading cause of work and activity absence. Prevention would not only save those affected from the burden of pain and discomfort, but would also save people from loss of over 290 million workdays annually and save the healthcare system billions of dollars in expenses per year. Successful research and development of a wearable technology capable of comprehensively monitoring spinal postures that are leading causes of back pain and injury can result in prevention of mild to severe back pain and injury for high-risk people. To accomplish this, the Smart Spine Tape is being developed with specific focus on accuracy, usability, and accessibility, all of which are important factors to consider when engineering for a wide array of populations. Accuracy was assessed using three human participants, with spinal angle data of the Smart Spine Tape being compared to established motion analysis technology data. Prototypes of the device showed promise in the ability to accurately measure spinal postures, but inconsistencies between samples and trials indicated that further development is necessary. Usability and accessibility were assessed using ten human participants who completed one workout each and reported on the tape’s comfort, durability, and ease of use, as well as their thoughts on how much they would be willing to pay for a fully functional version of the device. Participants reported high comfort, high durability, and moderate ease of use throughout their experiences, with the average price range that they would be willing to pay being between $25 and $75. Future directions have been identified that address inconsistencies in data collected by the Smart Spine Tape, possibly caused by inconsistent resistive properties of the piezoresistive ink and plastic deformation of the tape during testing. These future directions involve modifying testing, material, and fabrication methods.

Wearable technology offers the promise of decentralized and personalized healthcare, which can both alleviate current burdens on medical resources, and also help individuals to be more informed about their health. The heterogeneity of disease phenotypes necessitates adaptations to both diagnosing and surveilling disease, but to ensure user adoption and behavioral change, there needs to be a convenient way to amass such health information continuously. This can be in part accomplished by the development of continuously monitoring, compact wearable medical sensors and analytics technology that provide updates on analyte and biosignal measurements at regular intervals in situ. This dissertation investigates methods for collecting and analyzing information from wearable devices with these principles in mind. In Aim 1, we developed new methods for analysis of cardiovascular biosignals. Current methods of estimating left ventricular mass index (LVMI, a strong risk factor for cardiac outcomes), rely on the analysis of echocardiographic signals. Though still the gold standard, echocardiography can typically only be performed in the clinic, making it inconvenient to obtain frequent measurements of LVMI. Frequent measurements can be useful for monitoring cardiac risk, particularly for high-risk individuals, so we investigated the feasibility of predicting LVMI using a deep learning-based approach through ambulatory blood pressure readings, a one-time laboratory test and demographic information. We find that adding blood pressure waveform information in conjunction with multitask learning improved prediction errors (compared to baseline linear regression and neural network models), pointing to its potential as a clinical tool. Using transfer learning, we developed a model that does not require waveform data, but achieved similar prediction accuracies as methods that do require such data – opening the door to use cases that eliminate the need for wearing a blood pressure cuff continuously during the measurement period. Overall, such a technique has the potential to provide information to individuals who are at high risk of cardiac outcomes both inside and outside the clinic. In Aims 2 and 3, we developed a minimally invasive hydrogel patch for continuous monitoring of calcium, as proof-of-concept for wearable measurement of a wide variety of analytes typically assayed in the lab – a technology that can facilitate treatment and management of many prevalent diseases. Specifically, in Aim 2, we engineered a DNA polyacrylamide hydrogel microneedle array that sensed physiologically relevant calcium levels, for potential use by individuals who have hypoparathyroidism, a condition in which blood calcium levels are low and calcium supplements are needed. A negative mold was made using a CNC mill, the positive mold was cast in silicone, and the aptamer along with acrylamide and bis-acrylamide was seeded into the silicone mold. The DNA hydrogel was then fabricated using a simple UV curing protocol. The optimized DNA hydrogel was specific to calcium, used simple fabrication methods and had a fast, reversible signal response. Finally, in Aim 3, we developed the DNA hydrogel sensor into a wearable, integrated system with real-time fluorescence monitoring for testing in vivo. The microneedle array needed to be hydrated for the DNA aptamer to function, but polyacrylamide was too weak in its hydrated state to effectively pierce through skin epidermis. We demonstrated a method for strengthening our hydrogel system with polyethylene glycol diacrylate (PEGDA), while maintaining an optically translucent gel for detection purposes. We conducted piercing studies with a skin phantom on different microneedle array sizes and shapes, and determined that a 3x3 array of beveled microneedles required the least amount of force to pierce through a skin phantom. A custom complementary metal-oxide semiconductor (CMOS) system was developed to capture real-time fluorescence signals from the microneedle array, which correlated to calcium levels in vitro. This setup was then validated in a rat study. In this dissertation, we demonstrated methods for monitoring human biosignals using signal processing techniques, material innovations and integrated sensing platforms. While a work in progress, this dissertation is a step towards realizing the goal of decentralized, connected health for earlier detection and better management of disease.

Soft robotic sensors (SRSs) are a class of pliable, passive sensors which vary by some electrical characteristic in response to changes in geometry. The properties of SRSs make them excellent candidates for use in wearable motion analysis technology. Wearable technology is a fast-growing industry, and the improvement of existing human motion analysis tools is needed. Prior research has proven the viability of SRSs as a tool for capturing motion of the foot-ankle complex; this work covers extensive effort to improve and ruggedize a lab tool utilizing this technology. The improved lab tool is validated against a camera-based motion capture system to show either improvement or equivalence to the previous prototype while introducing enhanced data throughput, reliability, battery life, and durability.

Wearable tech has become increasingly popular with elite level sports organizations. The limiting factor to the value of the wearables is the use cases for the data they provide. This study introduces a technique to be used in tandem with this data to better inform training decisions. K-means clustering was used to group athletes from two seasons worth of data from an NCAA Division 1 American Football team. This data provided average game demands of each student-athlete, which was then used to create training groups. The resultant groupings showed results that were similar to traditional groupings used for training in American football, thus validating the results, while also offering insights on individuals that may need to consider training in a non-traditional group. In conclusion, this technique can be brought to athletic training and be useful in any organization that is dealing with training multitudes of athletes.

No description available.