2026-08-17: Visiting Research Position at Beth Israel Deaconess Medical Center | Harvard Medical School

In Summer 2026, I was honored to have the opportunity of being a visiting researcher at Beth Israel Deaconess Medical Center (BIDMC), a teaching hospital of Harvard Medical School in Boston. I worked at the Center of Education Research, Technology, and Innovation (CERTAIN) research center in the Department of Anesthesia, Critical Care, & Pain Medicine, under the supervision of Dr. Jacqueline Hannan.

CERTAIN research center focuses on educational research aimed at understanding and improving how healthcare professionals learn and train. CERTAIN combines educational research with emerging technologies to develop and evaluate new approaches to medical training and assessment. Its work includes both simulation-based training and clinical skills development. 

A particularly exciting aspect of the CERTAIN’s work is its use of realistic training environments. Residents participate in virtual reality activities, simulation-lab sessions, and in-situ training conducted directly in operating rooms. These experiences allow trainees to practice procedural skills, clinical decision making, and other aspects of patient care in realistic clinical scenarios while providing researchers with opportunities to study how clinicians learn and perform in these environments.

The research experience spanned for five months, from April through August. I spent the first three months working in person, primarily located at BIDMC East Campus, before continuing the projects remotely for the remainder of the summer. During this time, I worked closely with Dr. Hannan on analyzing gaze patterns and cognitive workload among medical residents in simulation-based training using eye-tracking data. 

During an In Situ simulation data collection in an operating
 room at BIDMC, with my supervisor, Dr. Hannan

One of the most valuable parts of the in-person experience was the opportunity to have daily check-ins with Dr. Hannan. We discussed my progress, reviewed findings, set goals for the day, and worked through questions as they came up. These regular interactions helped me better understand concepts from the medical domain and receive constant feedback as I worked through the research.

I also participated in weekly research meetings with the broader group and Dr. Alexander Shtifman, Administrative Director of Research and Graduate Medical Programs within the anesthesia department at BIDMC. Being part of these discussions gave me an opportunity to learn from researchers and clinicians from different backgrounds and to see how interdisciplinary perspectives can come together to address challenges in medical education and training.

Left: Beth Israel Deaconess Medical Center - East Campus, Right: Harvard Medical School

During my research experience, I worked on two projects with the focus of understanding how medical residents visually navigate and process complex medical procedures using eye-tracking measures.

Project 1: Simulated Ultrasound-Guided Central Venous Catheter Placement

In this project, we explored visual attention patterns and cognitive workload during a simulated ultrasound guided central venous catheter (CVC) placement task among novice anesthesia trainees using eye tracking data. Central venous catheter placement is an essential skill in anesthesia and involves inserting a catheter into a larger vein, often in the neck or chest. Because of the procedure’s technical complexity, it can be challenging for trainees to learn. Simulation-based training provides a controlled environment for evaluating procedural performance without putting patients at risk. 

For this project, novice anesthesiology residents at BIDMC completed a simulated ultrasound-guided internal jugular CVC placement using a CentralLineMan manikin task trainer. While they performed the procedure, we collected their eye-tracking data using Pupil Labs Neon wearable eye-tracking glasses. This device records binocular eye movements, pupillary measurements, and first-person view of the task. This allowed us to examine where they were looking and how their visual attention changed throughout the task.

Sample data collection session during a simulated ultrasound-guided central venous catheter placement task, performed by a member of CERTAIN research center, wearing the Pupil Labs Neon wearable eye-tracking glasses

We used Pupil Cloud software to extract eye tracking data including gaze positions, pupil measurements, fixation events, and saccades events. Fixations are the periods where the gaze remains stationary, and saccades are the rapid gaze shifts. Fixation and saccade events are identified by the fixation classification algorithm implemented in Pupil Cloud. This algorithm uses an extended Identification by Velocity-Threshold (I-VT) approach using adaptive velocity threshold.

Looking at Different Stages of the Procedure

A key part of the analysis was examining how visual attention and cognitive load changed as the procedure progressed. We divided the procedure into several meaningful procedural stages based on important events in the CVC placement process. 

This allowed us to ask whether visual attention and cognitive workload changed as trainees progressed through different stages of the procedure:

• Stage 1 - Preparation and Needle Insertion: From the start of the procedure to needle insertion.
• Stage 2 - Needle Guidance: From needle insertion to syringe removal for guidewire insertion
• Stage 3 - Procedure Completion: From syringe removal for guidewire insertion to the procedure completion

We also defined areas of interest (AOIs) using the image-mapping feature in Pupil Cloud software. We annotated the AOIs on a reference image of the experiment setup (see the Figure below) to identify task relevant objects; ultrasound screen, and manikin task trainer. This allowed us to examine how participants distributed their visual attention between these task objects throughout the procedure.

Annotated Areas of Interest (AOIs) in the experiment setup during the simulated central venous catheter placement task. The ultrasound is highlighted in red, and the manikin task trainer is highlighted in blue.

Combining Traditional and Advanced Gaze Measures

We analyzed a combination of traditional gaze measures and advanced gaze measures. Traditional gaze measures include pupillometry metric, fixation metrics (fixation duration and fixation count), and saccade metrics (saccade amplitude, saccade duration, and saccade velocity). We calculated advanced gaze metrics from the gaze and pupil data. 

One of the advanced gaze measures was the ambient/focal coefficient K, which captures dynamic shifts between broad exploratory visual behavior and focused inspection. We also examined, Low/High Index of Pupillary Activity (LHIPA), a luminance-robust pupillometry-based metric that estimates cognitive load. 

Together, these measures allowed us to analyze how participants’ visual attention and cognitive load changed during different stages of the task as well as how their focus were distributed between the task-relevant objects.

What We Learned

One of the interesting observations from this project was how visual attention changed during different stages of the procedure. The needle guidance phase, between needle insertion and guidewire preparation, stood out as a period of more focused visual processing. During this part of the procedure, trainees need to coordinate several things at once: manipulating the needle, monitoring the ultrasound image, identifying the needle tip, and coordinating their hand movements with what they see on the screen. This observation suggests that this critical portion of the procedure is associated with more focused visual processing, while other stages showed broader visual exploration.

Another key observation was that participants directed more of their visual attention toward the ultrasound monitor than the manikin task trainer throughout the procedure. This highlights the importance of the ultrasound image as trainees continuously interpret the ultrasound image while coordinating their hand movements and next procedural steps. 

Project 2: In-Situ Simulation

The CERTAIN research group conducts weekly in-situ simulation training sessions with anesthesia residents. In this project, we explored cognitive workload among the residents during one of these simulation training sessions, which took place in a real operating room. The session focused on a clinical scenario involving CO2 embolism occurring during a laparoscopic surgical procedure.

The simulation involved a multidisciplinary team consisting of the resident undergoing the training and other clinicians serving in simulated roles, including a surgeon and an anesthesiologist. A patient manikin was used to simulate the surgical procedure and associated clinical events. While they performed the procedure, we collected eye-tracking data from the resident using Pupil Labs Neon wearable eye-tracking glasses. Following the task, we collected resident self-reflection survey responses and NASA Task Load Index (NASA-TLX) data to capture their perceived workload and experience during the task. 

For this project, we focused primarily on pupil diameter and the resident’s first-person view of the task. We used Pupil Cloud to extract pupil measurements and corresponding timestamps. These data were used to compute objective measures of cognitive workload, which were subsequently analyzed in relation to the simulation task performance and the resident’s self-reported workload.

In-Situ simulation setup

Pupillometry-based Cognitive Workload Analysis

We explored pupillometry as an indicator of cognitive load of residents during the in-situ simulation task. In this study, we used two pupillometry-based measures: average pupil diameter and the Low/High Index of Pupillary Activity (LHIPA), a luminance-robust metric designed to capture changes in pupillary activity associated with cognitive processing.

We examined whether these measures varied across several factors related to participants and task demands. Specifically, we asked:

• Does average pupil diameter or LHIPA differ between junior and senior participants?
• Do these measures differ between observations with shorter and longer task durations?
• Do participants reporting lower versus higher perceived cognitive workload on NASA-TLX show differences in average pupil diameter or LHIPA values?

We also examined the relationship between the objective pupillometry measures and subjective NASA-TLX workload to determine whether physiological measures of cognitive workload associated with participants’ perceived workload. 

What We Learned

Although our statistical analyses, including group comparison and correlation analyses, did not reveal significant differences or relationships, the work provided a foundation for exploring pupillometry-based cognitive workload assessment in real-world, in-situ simulation environments. Rather than relying solely on subjective workload measures, such as NASA-TLX, this study allowed us to investigate how objective pupil-based measures may be used to capture changes in cognitive workload. It also provided an opportunity to evaluate the potential utility of advanced pupillometry metrics for understanding cognitive workload during complex and naturalistic tasks. 

Life in Boston

Beyond the research experience, living in Boston during one of the most beautiful times of the year was a memorable part of my visit. Boston was especially vibrant in April and May, when the city came alive with spring blossoms and greenery. One of the highlights of my time there was watching the Boston Marathon and experiencing the incredible energy and spirit that filled the city.

I also had the opportunity to explore the Harvard University campus, including Harvard Square and Harvard Yard, as well as Harvard Museum of Natural History. I enjoyed visiting historic sites such as the Boston Tea Party site and Boston Common and walking through the charming cobblestone streets of Beacon Hill. This historic architecture and picturesque neighborhoods made even an ordinary walk around the city feel special.

Highlights from my summer in Boston, exploring history, landmarks, and cuisines

Reflections and Takeways

This research experience gave me the opportunity to apply my background in eye tracking and data analysis to real-world medical simulation studies. Through projects examining visual attention and cognitive workload during clinical training, I gained hands-on experience working with eye-tracking data collected in realistic environments. I also learned to adapt data processing and analysis approaches to real-world conditions, interpret findings in the context of clinical procedures, and communicate technical concepts to clinical personnel. Most importantly, I was excited to see firsthand how eye-tracking technology and the method we develop can contribute to healthcare research and medical education.

This experience was also an opportunity to share the NIRDS Lab’s in-house eye-tracking pipelines and algorithms with the medical community and explore the utility of these tools beyond controlled laboratory studies. Working with data from a realistic clinical environment presented a challenge compared with controlled user studies. In the lab, we can design conditions with clear differences to validate our methods and measures. In real-world scenarios, however, gaze behavior is more dynamic, and there may not be a clearly defined “expected” gaze pattern for comparison. This experience encouraged me to think more about how our measures can be interpreted and applied in practical, real-world settings.

Beyond research, living in a historic city like Boston, exploring new places, meeting people from different backgrounds, and experiencing life in a new environment added another meaningful dimension to my visit. Overall, this experience strengthened my technical research skills, expanded my interdisciplinary perspective, and left me excited about the possibilities of applying eye-tracking technology to real-world challenges in healthcare and beyond.

Acknowledgement

I am extremely grateful to my supervisor, Dr. Jacqueline Hannan, for giving me this invaluable opportunity and for her mentorship, encouragement, and support throughout my research experience. Her guidance and efforts in helping me connect with researchers and expand my professional network made this experience especially meaningful. 

I would also like to thank Dr. Alexander Shtifman and the team at the CERTAIN Research Center for welcoming me into their group and for their support during the data collection sessions and throughout my research position.

I am thankful to my advisor, Dr. Sampath Jayarathna, for supporting me during this research position and for his guidance. My thanks also go to Dr. Dushan Wadduwage for helping me prepare for my visit to Boston and for his advice and encouragement in expanding my professional network. I am grateful to NIRDS Lab, WS-DL research group, and Computer Science department for their continued support in my graduate journey.

--Yasasi Abeysinghe (@Yasasi_Abey)