Vu Phan

Vu Phan

Auto-generated profile, based on publicly available information

Vu Phan

Carnegie Mellon University, Pittsburgh, PA, USA

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
21 February 2024 in IEEE Journal of Biomedical and Health Informatics
Verified
Journal Article
Closed access
VPVu Phan
Vu Phan
KSKe Song
Ke Song
RSRodrigo Scattone Silva
Rodrigo Scattone Silva
KSKarin Gravare Silbernagel
Karin Gravare Silbernagel
JBJosh Baxter
Josh Baxter
EHEni Halilaj
Eni Halilaj
Objective: Exercise monitoring with low-cost wearables could improve the efficacy of remote physicaltherapy prescriptions by tracking compliance and informing the delivery of tailored feedback. While a multitude of commercial wearables can detect activities of daily life, such as walking and running, they cannot accurately detect physical-therapy exercises. The goal of this study was to build open-source classifiers for remote physical therapy monitoring and provide insight on how data collection choices may impact classifier performance. Methods: We trained and evaluated multi-class classifiers using data from 19 healthy adults who performed 37 exercises while wearing 10 inertial measurement units on the wrist, pelvis, thighs, shanks, and feet. We investigated the effect of sensor density, location, type, sampling frequency, output granularity, feature engineering, and training-data size on exercise-classification performance. Results: Exercise groups (n = 10) could be classified with 96% accuracy using a set of 10 inertial measurement units (IMUs) and with 89% accuracy using a single pelvis-worn IMU. Multiple sensor modalities (i.e., accelerometers and gyroscopes), high sampling frequencies, and more data from the same population did not improve model performance, but in the future data from diverse populations and better feature engineering could. Conclusions: Given the growing demand for exercise monitoring systems, our sensitivity analyses, along with open-source tools and data, should reduce barriers for product developers, who are balancing accuracy with product formfactor, and increase transparency and trust in clinicians and patients. The open-source data and code are available at https://simtk.org/projects/imu-exercise.

15 October 2025 in IEEE Transactions on Biomedical Engineering
Verified
Journal Article
Open access
ZLZhixiong Li
Zhixiong Li
SSSoyong Shin
Soyong Shin
VPVu Phan
Vu Phan
EMEvy Meinders
Evy Meinders
EHEni Halilaj
Eni Halilaj
Objective: Markerless motion tracking using computer vision offers a scalable alternative to traditional marker-based analysis. While insight on how different emerging solutions compare could inform adoption across applications by highlighting accuracy-complexity tradeoffs, comprehensive benchmarking of these methods on the same open dataset remains a gap. This study evaluated thirteen single-view and two multi-view markerless motion capture methods against marker-based motion tracking for lower-extremity kinematics. Twenty-three healthy adults performed walking and lower-limb functional exercises, recorded with 20 infrared and 10 red-green-blue (RGB) cameras. The best-performing single-view model (WHAM) was used to test the hypothesis that the anatomical constraints imposed by biomechanical models improve the accuracy of vision-only solutions. We found that the addition of biomechanical modeling (bioWHAM) does not significantly improve overall kinematics accuracy, with root-mean-square differences (RMSD) sometimes being lower for the baseline model (WHAM) and sometimes for bioWHAM, with the median fluctuations across tasks being under 1.7$^\circ $, in either direction. Multi-view methods generally outperformed single-view ones: OpenCap, an open-source solution using two cameras, outperformed WHAM by 1.7$^\circ $ (p < 0.0001), whereas Theia3D, a commercial software using ten cameras, outperformed OpenCap by 1.3$^\circ $ (p < 0.0001). These findings suggest that while multi-view systems can enhance accuracy, the marginal benefits may not justify the added complexity across all applications. In addition to helping users consider the hardware, software, and accuracy tradeoffs, these findings highlight the need for continued innovation in multi-view fusion and incorporation of biomechanical modeling and computer vision. The accompanying dataset (I-MOVE-23) should facilitate continued benchmarking of emerging solutions.

27 April 2022 in IEEE Transactions on Biomedical Engineering
Verified
Journal Article
Closed access
VPVu Phan
Vu Phan
HLHyunglae Lee
Hyunglae Lee
Objective: This study introduces a deep learning approach to accurately predict challenging mechanical environments that possibly cause decreasing postural stability. Methods: Dual-axis robotic platforms were utilized to simulate various environments and collect center-of-pressure data during narrow and wide stance. A convolutional neural network (CNN) was developed to predict environmental conditions given segmented time-series balance data. Different window sizes were examined to investigate its minimal length for reliable prediction. Effectiveness of the presented CNN was additionally compared with that of conventional machine learning models. Its applicability with low sampled data or more natural stance data was then evaluated. Results: The CNN achieved above 94.5% in the overall prediction accuracy even with 2.5-second length postural sway data, which cannot be achieved by traditional machine learning ( ps < 0.05). Increasing data length beyond 2.5 seconds slightly improved the accuracy of CNN but substantially increased training time (60% longer). Importantly, results from averaged normalized confusion matrices revealed that CNN is much more capable of differentiating the mid-level environmental condition. Deep learning could also produce comparable performance in predicting environments even with much lower sampled data or with standing posture changed. Conclusion: CNN removed the burden of feature preparation and accurately predicted environments when dealing with short-length data. It also indicated potentials to real life applications. Significance: This study contributes to the advancement of wearable devices and human interactive robots (e.g., exoskeletons and prostheses) by predicting environmental contexts and preventing potential falls.

12 April 2023 in IEEE Transactions on Medical Robotics and Bionics
Verified
Journal Article
Closed access
CPConnor M. Phillips
Connor M. Phillips
VPVu Phan
Vu Phan
KJKwanghee Jo
Kwanghee Jo
OSOmik Save
Omik Save
JRJoshua B. Russell
Joshua B. Russell
HLHyunglae Lee
Hyunglae Lee
Children with cerebral palsy suffer from balance deficits that may greatly reduce their quality of life. However, recent advancements in robotics allow for balance rehabilitation paradigms that provide greater control of the training environment and more robust measurement techniques. Previous works have shown functional balance improvement using standing surface perturbations and compliant surface balancing. Visual feedback during balance training has also been shown to improve postural balance control. However, the combined effect of these interventions has not been evaluated. This paper presents a robotaided rehabilitation study for children with cerebral palsy on a side-specific performance-adaptive compliant surface with perturbations. Visual feedback of the participants center of pressure and weight distribution were used to evaluate successful balance and trigger perturbations after a period of successful balancing. The platform compliance increased relative to the amount of successful balance during each training interval. Two participants trained for 6 weeks including 10, less than 2 hours long, training sessions. Improvements in functional balance as assessed by the Pediatric Balance Scale, Timed 10 Meter Walk Test, and 5 Times Sit-to-Stand Test were observed for both participants. There was a reduction in fall risk as evidenced by increased Virtual Time to Contact and an increase in dynamic postural balance supported by a faster Time to Perturb, Time to Stabilize, and Percent Stabilized. A mixed improvement in static postural balance was also observed. This paper highlights the feasibility of robot-aided rehabilitation interventions as a method of balance therapy for children with cerebral palsy.

28 April 2023 in IEEE Journal of Translational Engineering in Health and Medicine
Verified
Journal Article
Open access
JRJoshua B. Russell
Joshua B. Russell
CPConnor M. Phillips
Connor M. Phillips
MAMatthew R. Auer
Matthew R. Auer
VPVu Phan
Vu Phan
KJKwanghee Jo
Kwanghee Jo
HLHyunglae Lee
Hyunglae Lee
Objective: This paper presents a twin dual-axis robotic platform system which is designed for the characterization of postural balance under various environmental conditions and quantification of bilateral ankle mechanics in 2 degrees-of-freedom (DOF) during standing and walking. Methods: Validation experiments were conducted to evaluate performance of the system 1) to apply accurate position perturbations under different loading conditions, 2) to simulate a range of stiffness-defined mechanical environments, and 3) to reliably quantify the joint impedance of mechanical systems. In addition, several human experiments were performed to demonstrate the system’s applicability for various lower limb biomechanics studies. The first two experiments quantified postural balance on a compliance-controlled surface (passive perturbations) and under oscillatory perturbations with various frequencies and amplitudes (active perturbations). The second two experiments quantified bilateral ankle mechanics, specifically, ankle impedance in 2-DOF during standing and walking. Results: The validation experiments showed high accuracy of the platform system to apply position perturbations, simulate a range of mechanical environments, and quantify the joint impedance. Results of the human experiments further demonstrated that the platform system is sensitive enough to detect differences in postural balance control under challenging environmental conditions as well as bilateral differences in 2-DOF ankle mechanics. Conclusion: This robotic platform system will allow us to better understand lower limb biomechanics during functional tasks, while also providing invaluable knowledge for the design and control of many robotic systems including robotic exoskeletons, prostheses and robot-assisted balance training programs.

7 June 2023 in TechRxiv
Verified
Preprint
Open access
PDF file
VPVu Phan
Vu Phan
KSKe Song
Ke Song
RSRodrigo Scattone Silva
Rodrigo Scattone Silva
KSKarin G. Silbernagel
Karin G. Silbernagel
JBJosh R. Baxter
Josh R. Baxter
EHEni Halilaj
Eni Halilaj
Objective: Exercise monitoring with low-cost wearables could improve the efficacy of remote physical-therapy prescriptions by tracking compliance and informing the delivery of tailored feedback. While a multitude of commercial wearables can detect activities of daily life, such as walking and running, they cannot accurately detect physical-therapy exercises. The goal of this study was to build open-source classifiers for remote physical therapy monitoring and provide insight on how data collection choices may impact classifier performance. Methods: We trained and evaluated multi-class classifiers using data from 19 healthy adults who performed 37 exercises while wearing 10 inertial measurement units on the wrist, pelvis, thighs, shanks, and feet. We investigated the effect of sensor density, location, type, sampling frequency, output granularity, feature engineering, and training-data size on exercise-classification performance. Results: Exercise groups (n=10) could be classified with 96% accuracy using a set of 10 inertial measurement units (IMUs) and with 89% accuracy using a single pelvis-worn IMU. Multiple sensor modalities (i.e., accelerometers and gyroscopes), high sampling frequencies, and more data from the same population did not improve model performance, but in the future data from diverse populations and better feature engineering could. Conclusions: Given the growing demand for exercise monitoring systems, our sensitivity analyses, along with open-source tools and data, should reduce barriers for product developers, who are balancing accuracy with product formfactor, and increase transparency and trust in clinicians and patients.