Alan Tian Tan

All Publications

• Self-Supervised Learning Improves Accuracy and Data Efficiency for IMU-Based Ground Reaction Force Estimation. IEEE transactions on bio-medical engineering Tan, T., Shull, P. B., Hicks, J. L., Uhlrich, S. D., Chaudhari, A. S. 2024; PP

  Abstract

  OBJECTIVE: Recent deep learning techniques hold promise to enable IMU-driven kinetic assessment; however, they require large extents of ground reaction force (GRF) data to serve as labels for supervised model training. We thus propose using existing self-supervised learning (SSL) techniques to leverage large IMU datasets to pre-train deep learning models, which can improve the accuracy and data efficiency of IMU-based GRF estimation.METHODS: We performed SSL by masking a random portion of the input IMU data and training a transformer model to reconstruct the masked portion. We systematically compared a series of masking ratios across three pre-training datasets that included real IMU data, synthetic IMU data, or a combination of the two. Finally, we built models that used pre-training and labeled data to estimate GRF during three prediction tasks: overground walking, treadmill walking, and drop landing.RESULTS: When using the same amount of labeled data, SSL pre-training significantly improved the accuracy of 3-axis GRF estimation during walking compared to baseline models trained by conventional supervised learning. Fine-tuning SSL model with 1-10% of walking data yielded comparable accuracy to training baseline model with 100% of walking data. The optimal masking ratio for SSL is 6.25-12.5%.CONCLUSION: SSL leveraged large real and synthetic IMU datasets to increase the accuracy and data efficiency of deep-learning-based GRF estimation, reducing the need for labeled data.SIGNIFICANCE: This work, with its open-source code and models, may unlock broader use cases of IMU-driven kinetic assessment by mitigating the scarcity of GRF measurements in practical applications.

  View details for DOI 10.1109/TBME.2024.3361888

  View details for PubMedID 38315597

• A scoping review of portable sensing for out-of-lab anterior cruciate ligament injury prevention and rehabilitation. NPJ digital medicine Tan, T., Gatti, A. A., Fan, B., Shea, K. G., Sherman, S. L., Uhlrich, S. D., Hicks, J. L., Delp, S. L., Shull, P. B., Chaudhari, A. S. 2023; 6 (1): 46

  Abstract

  Anterior cruciate ligament (ACL) injury and ACL reconstruction (ACLR) surgery are common. Laboratory-based biomechanical assessment can evaluate ACL injury risk and rehabilitation progress after ACLR; however, lab-based measurements are expensive and inaccessible to most people. Portable sensors such as wearables and cameras can be deployed during sporting activities, in clinics, and in patient homes. Although many portable sensing approaches have demonstrated promising results during various assessments related to ACL injury, they have not yet been widely adopted as tools for out-of-lab assessment. The purpose of this review is to summarize research on out-of-lab portable sensing applied to ACL and ACLR and offer our perspectives on new opportunities for future research and development. We identified 49 original research articles on out-of-lab ACL-related assessment; the most common sensing modalities were inertial measurement units, depth cameras, and RGB cameras. The studies combined portable sensors with direct feature extraction, physics-based modeling, or machine learning to estimate a range of biomechanical parameters (e.g., knee kinematics and kinetics) during jump-landing tasks, cutting, squats, and gait. Many of the reviewed studies depict proof-of-concept methods for potential future clinical applications including ACL injury risk screening, injury prevention training, and rehabilitation assessment. By synthesizing these results, we describe important opportunities that exist for clinical validation of existing approaches, using sophisticated modeling techniques, standardization of data collection, and creation of large benchmark datasets. If successful, these advances will enable widespread use of portable-sensing approaches to identify ACL injury risk factors, mitigate high-risk movements prior to injury, and optimize rehabilitation paradigms.

  View details for DOI 10.1038/s41746-023-00782-2

  View details for PubMedID 36934194

• IMU and Smartphone Camera Fusion for Knee Adduction and Knee Flexion Moment Estimation During Walking IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS Tan, T., Wang, D., Shull, P. B., Halilaj, E. 2023; 19 (2): 1445-1455

  View details for DOI 10.1109/TII.2022.3189648

  View details for Web of Science ID 000926964700031

• Effects of IMU Sensor-to-Segment Misalignment and Orientation Error on 3-D Knee Joint Angle Estimation IEEE SENSORS JOURNAL Fan, B., Li, Q., Tan, T., Kang, P., Shull, P. B. 2022; 22 (3): 2543-2552

  View details for DOI 10.1109/JSEN.2021.3137305

  View details for Web of Science ID 000750206700082

• Transfer Learning Improves Accelerometer-Based Child Activity Recognition via Subject-Independent Adult-Domain Adaption. IEEE journal of biomedical and health informatics Li, J., Kang, P., Tan, T., Shull, P. B. 2021; PP

  Abstract

  Wearable activity recognition can collate the type, intensity, and duration of each childs physical activity profile, which is important for exploring underlying adolescent health mechanisms. Traditional machine-learning-based approaches require large labeled data sets; however, child activity data sets are typically small and insufficient. Thus, we proposed a transfer learning approach that adapts adult-domain data to train a high-fidelity, subject-independent model for child activity recognition. Twenty children and twenty adults wore an accelerometer wristband while performing walking, running, sitting, and rope skipping activities. Activity classification accuracy was determined via the traditional machine learning approach without transfer learning and with the proposed subject-independent transfer learning approach. Results showed that transfer learning increased classification accuracy to 91.4% as compared to 80.6% without transfer learning. These results suggest that subject-independent transfer learning can improve accuracy and potentially reduce the size of the required child data sets to enable physical activity monitoring systems to be adopted more widely, quickly, and economically for children and provide deeper insights into injury prevention and health promotion strategies.

  View details for DOI 10.1109/JBHI.2021.3118717

  View details for PubMedID 34623286

• Accurate Impact Loading Rate Estimation During Running via a Subject-Independent Convolutional Neural Network Model and Optimal IMU Placement IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS Tan, T., Strout, Z. A., Shull, P. B. 2021; 25 (4): 1215-1222

  Abstract

  Enable accurate estimation of vertical average loading rate (VALR) in runners with one or more wearable inertial measurement units (IMUs).A subject-independent convolutional neural network (CNN) model was developed to estimate VALR from wearable IMUs. Fifteen runners wore IMUs at the trunk, pelvis, thigh, shank, and foot and ran on an instrumented treadmill for combinations of the following conditions: foot-strike (forefoot, mid-foot, rear-foot), step rate (90% to 110% of baseline), running speed (2.4 m/s and 2.8 m/s) and footwear (standard and minimalist running shoes). Thirty-one IMU placement configurations with combinations of one to five IMUs were evaluated. VALR estimations from the wearable IMUs were compared with force-plate VALR measurements.VALR estimations via the subject-independent CNN model with a single shank-worn IMU were highly correlated (ρ = 0.94) with force-plate VALR measurements and were substantially higher than previously reported peak tibial acceleration correlations with force-plate VALR measurements from shank-worn accelerometers (ρ = 0.44-0.66). Correlation results from the CNN model for a single IMU placed at the foot, pelvis, trunk, and thigh were ρ = 0.91, 0.76, 0.69, and 0.65, respectively. There was no improvement in accuracy from the shank-worn IMU when adding 1-4 additional IMUs from the trunk, pelvis, thigh, or foot.The proposed subject-independent CNN model with a single shank-worn IMU provides more accurate estimation of VALR than previous wearable sensing approaches.This could enable runners to more accurately assess impact loading rates and potentially provide insights into running-related injury risk and prevention.

  View details for DOI 10.1109/JBHI.2020.3014963

  View details for Web of Science ID 000638401400031

  View details for PubMedID 32763858

• Magnetometer-Free, IMU-Based Foot Progression Angle Estimation for Real-Life Walking Conditions IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING Tan, T., Strout, Z. A., Xia, H., Orban, M., Shull, P. B. 2021; 29: 282-289

  Abstract

  Foot progression angle (FPA) is vital in many disease assessment and rehabilitation applications, however previous magneto-IMU-based FPA estimation algorithms can be prone to magnetic distortion and inaccuracies after walking starts and turns. This paper presents a foot-worn IMU-based FPA estimation algorithm comprised of three key components: orientation estimation, acceleration transformation, and FPA estimation via peak foot deceleration. Twelve healthy subjects performed two walking experiments to evaluation IMU algorithm performance. The first experiment aimed to validate the proposed algorithm in continuous straight walking tasks across seven FPA gait patterns (large toe-in, medium toe-in, small toe-in, normal, small toe-out, medium toe-out, and large toe-out). The second experiment was performed to evaluate the proposed FPA algorithm for steps after walking starts and turns. Results showed that FPA estimations from the IMU-based algorithm closely followed marker-based system measurements with an overall mean absolute error of 3.1±1.3 deg, and the estimation results were valid for all steps immediately after walking starts and turns. This work could enable FPA assessment in environments where magnetic distortion is present due to ferrous metal structures and electrical equipment, or in real-life walking conditions when walking starts, stops, and turns commonly occur.

  View details for DOI 10.1109/TNSRE.2020.3047402

  View details for Web of Science ID 000626331500008

  View details for PubMedID 33360997

• Influence of IMU position and orientation placement errors on ground reaction force estimation JOURNAL OF BIOMECHANICS Tan, T., Chiasson, D. P., Hu, H., Shull, P. B. 2019; 97: 109416

  Abstract

  Wearable inertial measurement units (IMU) have been proposed to estimate GRF outside of specialized laboratories, however the precise influence of sensor placement error on accuracy is unknown. We investigated the influence of IMU position and orientation placement errors on GRF estimation accuracy.Kinematic data from twelve healthy subjects based on marker trajectories were used to simulate 1848 combinations of sensor position placement errors (range ± 100 mm) and orientation placement errors (range ± 25°) across eight body segments (trunk, pelvis, left/right thighs, left/right shanks, and left/right feet) during normal walking trials for baseline cases when a single sensor was misplaced and for the extreme cases when all sensors were simultaneously misplaced. Three machine learning algorithms were used to estimate GRF for each placement error condition and compared with the no placement error condition to evaluate performance.Position placement errors for a single misplaced IMU reduced vertical GRF (VGRF), medio-lateral GRF (MLGRF), and anterior-posterior GRF (APGRF) estimation accuracy by up to 1.1%, 2.0%, and 0.9%, respectively and for all eight simultaneously misplaced IMUs by up to 4.9%, 6.0%, and 4.3%, respectively. Orientation placement errors for a single misplaced IMU reduced VGRF, MLGRF, and APGRF estimation accuracy by up to 4.8%, 7.3%, and 1.5%, respectively and for all eight simultaneously misplaced IMUs by up to 20.8%, 23.4%, and 12.3%, respectively.IMU sensor misplacement, particularly orientation placement errors, can significantly reduce GRF estimation accuracy and thus measures should be taken to account for placement errors in implementations of GRF estimation via wearable IMUs.

  View details for DOI 10.1016/j.jbiomech.2019.109416

  View details for Web of Science ID 000502890000020

  View details for PubMedID 31630774

• Resonant Frequency Skin Stretch for Wearable Haptics. IEEE transactions on haptics Shull, P. B., Tan, T., Culbertson, H. M., Zhu, X., Okamura, A. 2019

  Abstract

  Resonant frequency skin stretch uses cyclic lateral skin stretches matching the skin's resonant frequency to create highly noticeable stimuli, signifying a new approach for wearable haptic stimulation. Three experiments were performed to explore biomechanical and perceptual aspects of resonant frequency skin stretch. In the first experiment, effective skin resonant frequencies were quantified at the forearm, shank, and foot. In the second experiment, perceived haptic stimuli were characterized for skin stretch actuations across a spectrum of frequencies. In the third experiment, haptic classification ability was determined as subjects differentiated haptic stimulation cues while sitting, walking, and jogging. Results showed that subjects perceived stimulations at, above, and below the skin's resonant frequency differently: stimulations lower than the skin resonant frequency felt like distinct impacts, stimulations at the skin resonant frequency felt like cyclic skin stretches, and stimulations higher than the skin resonant frequency felt like standard vibrations. Subjects successfully classified stimulations while sitting, walking, and jogging, and classification accuracy decreased with increasing speed, especially for stimulations at the shank. This work could facilitate more widespread use of wearable skin stretch. Potential applications include gaming, medical simulation, and surgical augmentation, and for training to reduce injury risk or improve sports performance.

  View details for DOI 10.1109/TOH.2019.2917072

  View details for PubMedID 31095499