M.Eng. Graduates

Jonah Ayala, Cameron “Cami” Chan, Yiwen Chen, Angely Morales-Nunez, Arena Rahman

Transcatheter aortic valve replacement (TAVR) has significantly improved outcomes for patients with aortic stenosis by providing a less invasive alternative to open-heart surgery. However, current TAVR systems face critical limitations, including compromised coronary access and limited long-term durability due to structural valve deterioration, often caused by calcification within ten years. To address these limitations, we propose a modular TAVR system with a collapsible stent design that is retrieved using a specialized transfemoral catheter. This design aims to preserve coronary access, reduce paravalvular leakage, and maintain optimal hemodynamic performance. Our approach allows for iterative re-interventions where both delivery and full removal occurs via catheter transport. This potentially eliminates the need for invasive, open-heart surgery to remove or replace a TAVR implant, ultimately enabling safer lifetime valve management.

Anjali Katta, Ashraf Alnatour, Hank Wang, Rick Han

Effective pain assessment in medical-surgical units remains challenging due to subjectivity, inconsistent monitoring, and limited clinician availability. We present an autonomous pain assessment and monitoring system that continuously evaluates patient pain levels using a multimodal imaging and machine learning approach. A custom-designed triple camera hardware platform captures synchronized patient images from multiple optimized angles, ensuring robust facial visibility despite occlusions or patient movement. These images are processed by a trained deep learning model capable of detecting and classifying facial action units associated with pain expression, generating a composite pain score ranging from 1 to 15. The resulting data are transmitted to a centralized dashboard accessible to nursing staff, enabling real-time monitoring across multiple patients. The interface provides longitudinal pain trends over the previous 24 hours, alongside corresponding visual data and medication administration history, supporting more informed and timely clinical decision-making. By automating pain detection and integrating it with clinical workflows, this system reduces reliance on intermittent manual assessments and enhances patient care consistency. This approach demonstrates the potential for scalable, non-invasive pain monitoring solutions that improve clinical responsiveness, optimize treatment strategies, and ultimately contribute to better patient outcomes in hospital settings.

Lauren Howe-Meagher, Jingzhe Li, Mingze Li, Faith Nwando, Shixing Yuan

Medical-surgical (med-surg) hospital wards rely on intermittent, manually collected vital signs, typically recorded every 4 to 8 hours. This approach creates gaps in monitoring during which clinically significant deterioration often goes undetected, contributing to delayed intervention, increased ICU transfers, and added strain on nursing staff. Routine vital sign collection also disrupts patient rest and increases workflow burden, particularly during overnight hours. This project presents the design of a non-contact, continuous vital sign monitoring system aimed at improving early detection of physiological decline while reducing reliance on manual measurements. A comprehensive analysis of disease-state fundamentals, current monitoring technologies, market drivers, and stakeholder needs informed the development process. Key design priorities included med-surg compatibility, compact form factor, continuous monitoring capability, and minimal non-critical disruptions to patients and care workflows. The team’s proposed system aims to enhance early-warning capabilities, reduce nursing workload, and improve patient comfort, supporting safer and more efficient inpatient care. The team’s proposed system leverages non-contact sensing technologies, including millimeter-wave radar, to enable continuous monitoring without restrictive wiring or direct patient contact.

Anvitha Doddipalli, Damilare Femi-Akanbi, Joshua Lee, Alton Shen, Jacqueline Walker MPH

This project presents the early-stage development of OsteoIQ, a novel non-invasive bone density measurement device to improve screening for osteopenia and osteoporosis. Osteoporosis disproportionately affects older adults, particularly postmenopausal women. Current solutions are costly, require specialized equipment, and administered infrequently, limiting opportunities for prevention, timely intervention, or monitoring of disease progression or treatment response. Early screening is critical to preventing fractures and long-term morbidity.

This work focuses on establishing proof-of-concept for a multimodal sensor combining bioimpedance and quantitative ultrasound. Rather than revalidating the physiological basis of these modalities, this project evaluates translational feasibility within practical constraints. Emphasis is placed on reliably acquiring and interpreting signals at the distal radius, accounting for anatomical variability, geometric limitations, signal noise, and the need for reproducible, clinically relevant measurements.

Development includes two prototypes: a functional benchtop system for validating signal acquisition and processing, and an ergonomic (“looks-like, feels-like”) design to explore usability and form factor. To support validation, forearm analog models with controlled bone density levels were created using demineralized bone samples embedded in tissue-mimicking materials, enabling correlation of measured signals with known bone values. Preliminary results demonstrate the potential to extract meaningful features associated with bone density and support early algorithm development.

Sujin Chung, Yashvi Gupta, Jinghao Shen, Jiaxi Wang

Needle phobia affects over 63% of patients requiring injectable drug administration, reducing treatment adherence, and increasing procedural anxiety. Current injection devices inadequately minimize pain and insertion-related discomfort, representing a significant unmet need in drug delivery. Rather than assuming pain can be eliminated entirely, this project focused on reducing injection discomfort through a mechanical, user-friendly design compatible with existing clinical and self-administration workflows. Background research linked pain perception to measurable engineering factors, including needle geometry, insertion force, and tissue interaction. Through iterative concept screening against feasibility, safety, manufacturability, and regulatory requirements, we pursued two complementary design directions: optimizing needle bevel geometry and integrating ultrasonic vibration assistance. ANSYS simulation and bench testing using a polyurethane skin phantom confirmed that a 7-bevel tip reduces tissue stress at penetration, while vibration-assisted insertion meaningfully lowered peak insertion force. The final concept combines a 7-bevel needle tip with an ultrasonic transducer injector featuring a replaceable drug cartridge, targeting reduced penetration force and improved comfort for patients requiring frequent subcutaneous injections.

Jingxiang Huang, Hari Kohli, Odyssefs Liagkas, Luciana Ruggiero

Total knee arthroplasty (TKA) is among the most performed orthopedic procedures worldwide, yet 10–20% of patients report dissatisfaction postoperatively, largely due to soft-tissue imbalance affecting the collateral ligaments and patellofemoral mechanism. Current intraoperative assessment relies on subjective tactile feedback, introducing significant surgeon-to-surgeon variability and leaving patellar tendon tension largely unquantified. To address this, a modular tensiometer system was developed comprising two complementary sensing components: a buckle transducer assembly for real-time medial and lateral collateral ligament tension measurement, and a sensing patellar button for intraoperative patellofemoral pressure quantification. The buckle transducer integrates a SingleTact thin-film force sensor with stepped-rod frame, enabling tissue tension derivation across a range of ligament sizes. The patellar component replaces the standard trial button to capture contact forces during knee flexion. Both subsystems feed into a custom MATLAB acquisition and balance-assessment interface providing real-time, clinically interpretable feedback. Validation was conducted using artificial latex ligaments and a 6-degree-of-freedom robotic arm simulating intraoperative kinematics. The system achieved sub-Newton measurement precision, sufficient for detecting clinically meaningful tension asymmetries and supporting objective, reproducible soft-tissue balancing during TKA.

Haram Kim, Adarsh Singh, Divakar Badal (advised by Dr. Ilana L. Brito)

Bacterial extracellular vesicles (BEVs) play an important role in facilitating host-microbe interaction. Specifically, BEVs mediate cellular communication through the horizontal transfer of genetic material, virulence factors, and enzymes. By leveraging these innate communication pathways, BEVs can be engineered to encapsulate specific therapeutic cargo. Previous studies have shown BEVs as a potential delivery mechanism for gene editing tools to mammalian cells. Despite their potential, engineering BEVs for the efficient packaging and delivery of editing cargo to diverse bacterial species remains a significant limitation. In this work, we established E. coli-derived BEVs as a platform for protein cargo packaging and developed screening methods to quantify bacteria-BEV fusion rate. Using signal peptides, we successfully packaged functional proteins, including sfGFP, Cre recombinase, and transposases, and further demonstrate that they remain active. We were also able to transfer plasmid cargo to E. coli cells. Our findings highlight BEVs as a programmable vector for transkingdom delivery of gene-editing tools.

Pang-Wei Liu (advised by Dr. Yadong Wang)

This research seeks to enhance the manufacturing and processing of imidazole-based polymers for use in metallo-elastomer vascular grafts. Three primary objectives are pursued to mitigate existing constraints in cost, synthetic complexity, and material performance. Initially, alternative monomer protection mechanisms are investigated to optimize synthesis and decrease manufacturing expenses. Second, simpler analogs of imidazole-based polymers are developed and assessed by a one-step monomer synthesis method. Third, the electrowriting behavior of imidazole-based polymers in the presence of metal ions is extensively examined, emphasizing the enhancement of fiber stability and fusion in vascular grafts. This research offers insights into effective polymer design and processing techniques, aiding the advancement of sophisticated materials for vascular tissue engineering applications.

Yonglin Zhang (advised by Dr. Iwijn De Vlaminck)

Antibodies and T-cell receptors are the immune system’s central recognition machinery and a cornerstone of modern biologics, from monoclonal antibody therapeutics to engineered T-cell therapies. Wet-lab discovery of viable paired heavy/light-chain antibodies (VH + VL) costs roughly $2.6 billion and 10+ years per drug, motivating in silico generation of clinically relevant candidates. This project develops a deep generative pipeline for paired antibody sequences using two complementary diffusion paradigms, a continuous-latent model (LD4LG) and a discrete masked-token model (DPLM), both trained from scratch on 2.17M paired human B-cell receptor sequences curated from the Observed Antibody Space (OAS) and Observed TCR Space (OTS), with stringent MMseqs2 similarity reduction (95% / 90%) ensuring zero train–test leakage. Generated antibodies are conditioned on three biologically meaningful labels, isotype, V-gene family, and light-chain locus, and evaluated for structural plausibility (HMMER domain match, IgFold foldability), sequence diversity, and conditional fidelity. We identify a clear quality–diversity Pareto frontier between the two paradigms, providing a reusable template for therapeutic antibody design and future extension to paired-TCR generation for adoptive cell therapy