Jennifer Yu

Jennifer Yu

Auto-generated profile, based on publicly available information

Jennifer Yu

All from Memorial Sloan Kettering Cancer Center

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

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

Representative Publications

View All Publications
28 April 2023 in IEEE Sensors Journal
Verified
Journal Article
Closed access
HAHajar Abedi
Hajar Abedi
MMMartin Ma
Martin Ma
JHJames He
James He
JYJennifer Yu
Jennifer Yu
AAAhmad Ansariyan
Ahmad Ansariyan
George ShakerGeorge Shaker
George Shaker
In this paper, we use a multiple-input multiple-output (MIMO) radar for in-vehicle passenger detection. We propose a 2D CNN-LSTM (Convolutional Neural Network- Long Short-Term Memory) to accurately detect, count and classify passengers inside five-seater vehicles. Our deep learning model first extracts feature using a CNN model, and then a series of frames will be delivered to a time series model (LSTM) to predict new scenarios. Additionally, we provide the outcomes of various deep learning models and show that temporal deep learning models perform better in our radar datasets. Furthermore, we provide reliable session-dependent datasets collected from different car models with various passengers (including infants/ children and adults). The results show that our proposed 2D CNN-LSTM model can detect unattended infants/ children in vehicles with more than 95% accuracy, count passengers and identify their occupied seats with an accuracy of 89% and classify passengers with more than 74% accuracy. Since our model is evaluated in a new car with new passengers, it ensures the generality of our proposed method to be deployed in any five-seater vehicle.

20 October 2025 in JMIR Formative Research
Verified
Journal Article
Open access
TBTina Behrouzi
Tina Behrouzi
JYJennifer Yu
Jennifer Yu
RYRobin Yang
Robin Yang
ABAdrien Boch
Adrien Boch
AGAnna Goldenberg
Anna Goldenberg
SFStephen H. Friend
Stephen H. Friend
Personal digital health technologies (DHTs) enable real-time monitoring of physiological metrics and behavioral data, including heart rate variability (HRV), supporting analysis of pregnancy-related conditions and personalized care throughout the perinatal period. While recent studies demonstrate the utility of personal DHTs in tracking pregnancy-related symptoms, they often rely on aggregate statistical methods that overlook individual variability. This study aims to compare aggregate and individual-level analyses of DHT data for pregnancy-related conditions, using the comprehensive BUMP (Better Understanding the Metamorphosis of Pregnancy) dataset to highlight the importance of individual variability and data heterogeneity. We analyzed wearable and self-reported data from 256 participants enrolled in the BUMP study (January 2021 to May 2022), including HRV, sleep, and fatigue measured via Oura Rings and smartphone surveys. Individual-level (N-of-1) trajectories were evaluated and compared with aggregate results to uncover personal and collective trends. A statistical method was developed to assess the influence of adverse events and severe symptoms, while case studies explored confounding and modifying factors underlying heterogeneity. Comprehensive statistical analysis included the coefficient of determination, Kolmogorov-Smirnov tests, likelihood ratio tests, and Welch t tests, with interindividual variability flagged based on high-variability thresholds. Substantial interindividual variability was observed across all features. Only 4.76% (12/256) of participants exhibited an HRV inflection at the aggregate week-33 inflection point, with a coefficient of variation of 14.24%. The median value of the gestational week in individual fatigue troughs was 23 (IQR 8; range 8-38) weeks, differing from aggregate estimates. Distributional comparisons showed no statistically significant differences in individual-level model fit (R²) by pregnancy complications or age (P values ranging from .06 to .99 across all model fit comparisons). Case studies further highlighted both intraindividual and interindividual differences, emphasizing the importance of considering external factors, such as adverse events and severe symptoms. Our findings show that aggregate wearable data often fail to generalize across populations, oversimplifying pregnancy-related physiological and subjective changes. This simplification can obscure individual trajectories, leading to generalized insights that may not reflect many pregnant women's experiences. Our results highlight the impact of heterogeneity on pregnancy outcomes, emphasizing the need to move beyond one-size-fits-all models and leverage DHT for personalized care.

22 October 2025 in medRxiv
Verified
Preprint
Open access
PDF file
TBTina Behrouzi
Tina Behrouzi
JYJennifer Yu
Jennifer Yu
RYRobin Yang
Robin Yang
ABAdrien Boch
Adrien Boch
AGAnna Goldenberg
Anna Goldenberg
SFStephen H. Friend
Stephen H. Friend
Background: Personal digital health technologies (DHTs) enable real-time monitoring of physiological metrics and behavioral data, including HRV, supporting early detection of pregnancy-related conditions and personalized care throughout the perinatal period. While recent studies demonstrate the utility of personal DHTs in tracking pregnancy-related symptoms, they often rely on aggregate statistical methods that overlook individual variability. Objective: To compare aggregate and individual-level analyses of digital health technology (DHT) data for early detection of pregnancy-related conditions, using the comprehensive BUMP dataset to highlight the importance of individual variability and data heterogeneity. Methods: This BUMP study (Jan 2021 – May 2022) analyzed physiological and behavioral metrics, such as heart rate variability (HRV), sleep, and fatigue, in 256 individuals using Oura rings and self-reported surveys. Individual-level (N-of-1) trajectories were evaluated and compared with aggregate results to uncover personal and collective trends. A statistical method was developed to assess the influence of adverse events and severe symptoms, while case studies explored confounding and modifying factors underlying heterogeneity. Comprehensive statistical analysis included the coefficient of determination, Kolmogorov-Smirnov tests, likelihood ratio tests, and Welch’s t-tests, with inter-individual variability flagged based on high-variability thresholds. Results: Results revealed significant variability in HRV, sleep, and fatigue throughout pregnancy. For instance, only 4.76% of individuals had HRV inflection points at the aggregate week 33 inflection, with a 14.24% coefficient of variation. Our analysis found no significant p-values for demographic or pregnancy complication-based subgrouping, suggesting these factors alone do not drive the observed variability. Case studies further highlighted both intra- and inter-individual differences, emphasizing the importance of considering external factors like adverse events and severe symptoms. Conclusions: Our findings show that aggregate wearable data often fails to generalize across populations, oversimplifying pregnancy-related physiological and subjective changes. This simplification can obscure individual trajectories, leading to generalized insights that may not reflect many pregnant women’s experiences. Our results highlight the impact of heterogeneity on pregnancy outcomes, emphasizing the need to move beyond one-size-fits-all models and leverage DHT for personalized care.

27 April 2024 in medRxiv
Verified
Preprint
Open access
PDF file
SGSarah Goodday
Sarah Goodday
RYRobin Yang
Robin Yang
EKEmma Karlin
Emma Karlin
JTJonell Tempero
Jonell Tempero
CHChristiana Harry
Christiana Harry
SFStephen Friend
Stephen Friend
Wearables, apps and other remote smart devices can capture rich, objective physiologic, metabolic, and behavioral information that is particularly relevant to pregnancy. The objectives of this paper were to 1) characterize individual level pregnancy self-reported symptoms and objective features from wearables compared to the aggregate; 2) determine whether pregnancy self-reported symptoms and objective features can differentiate pregnancy-related conditions; and 3) describe associations between self-reported symptoms and objective features. Data are from the Better Understanding the Metamorphosis of Pregnancy study, which followed individuals from preconception to three-months postpartum. Participants (18-40 years) were provided with an Oura smart ring, a Garmin smartwatch, and a Bodyport Cardiac Scale. They also used a study smartphone app with surveys and tasks to measure symptoms. Analyses included descriptive spaghetti plots for both individual-level data and cohort averages for select weekly reported symptoms and objective measures from wearables. This data was further stratified by pregnancy-related clinical conditions such as preeclampsia and preterm birth. Mean Spearman correlations between pairs of self-reported symptoms and objective features were estimated. Self-reported symptoms and objective features during pregnancy were highly heterogeneous between individuals. While some aggregate trends were notable, including an inflection in heart rate variability approximately eight weeks prior to delivery, these average trends were highly variable at the n-of-1 level, even among healthy individuals. Pregnancy conditions were not well differentiated by objective features. With the exception of self-reported swelling and body fluid volume, self-reported symptoms and objective features were weakly correlated (mean Spearman correlations <0.1). High heterogeneity and complexities of associations between subjective experiences and objective features across individuals pose challenges for researchers and highlights the dangers in reliance on aggregate approaches in the use of wearable data in pregnant individuals. Innovation in machine learning and AI approaches at the n-of-1 level could help to accelerate the field. Author Summary The objective physiological and behavioral information from wearable and other smart devices is uniquely relevant to pregnancy. The objectives of this study were to: 1) describe the individual-level variability of pregnancy self-reported symptoms and objective wearable measures; 2) determine whether this variability can be explained by pregnancy clinical conditions; and 3) determine whether pregnancy self-reported symptoms are associated with objective wearable measures. Data are from the Better Understanding the Metamorphosis of Pregnancy study, which followed individuals from preconception to three-months postpartum. Participants (18-40 years) used an Oura smartring, a Garmin smartwatch, and a Bodyport Cardiac Scale alongside a study app to track self-reported symptoms. High heterogeneity was observed in self-reported pregnancy symptoms, and objective measures such as heart rate variability, activity and sleep over pregnancy that were dissimilar to the population average of these measures. Pregnancy clinical conditions did not explain well the observed high variability in objective wearable measures while self-reported symptoms were weakly correlated with objective wearable measures over pregnancy. In sum, high heterogeneity and complexities of associations between subjective experiences and objective measures from wearables across pregnant individuals pose challenges for researchers. Innovation in machine learning and AI individual level approaches will help to accelerate the field.

22 May 2026 in bioRxiv
Verified
Preprint
Open access
PDF file
JYJennifer Yu
Jennifer Yu
MDMadison Darmofal
Madison Darmofal
MWMichele Waters
Michele Waters
JCJanuary Choy
January Choy
TTThinh N. Tran
Thinh N. Tran
JJJustin Jee
Justin Jee
General-purpose large language models (LLMs) are trained on large corpora to acquire broad knowledge, but whether LLMs can replace, or augment, task-specific models is unclear. We evaluated LLMs on three real-world, clinically important tumor genomic interpretation tasks, in order of increasing difficulty: (i) distinguishing tumor from non-tumor mutations (n=34,415 variants), (ii) distinguishing driver from passenger mutations (n=13,469 variants), and (iii) inferring cancer type from tumor sequencing reports across multiple assays and institutions (n=102,791 samples). The best general-purpose LLMs performed as well as the benchmark tailor-made predictor for task (i). Ensembling tailor-made models with zero-shot LLMs improved their performance for tasks (i) and (ii). For task (iii), LLMs outperformed or supplemented tailor-made models on out-of-distribution data. Without fine-tuning, current LLMs already can be useful in clinical genomic interpretation by adding complementary expertise to tailor-made, state-of-the-art predictors.

20 October 2025 in JMIR Preprints
Verified
Preprint
Open access
TBTina Behrouzi
Tina Behrouzi
JYJennifer Yu
Jennifer Yu
RYRobin Yang
Robin Yang
ABAdrien Boch
Adrien Boch
AGAnna Goldenberg
Anna Goldenberg
SFStephen H. Friend
Stephen H. Friend
BACKGROUND Personal digital health technologies (DHTs) enable real-time monitoring of physiological metrics and behavioral data, including HRV, supporting early detection of pregnancy-related conditions and personalized care throughout the perinatal period. While recent studies demonstrate the utility of personal DHTs in tracking pregnancy-related symptoms, they often rely on aggregate statistical methods that overlook individual variability. OBJECTIVE To compare aggregate and individual-level analyses of digital health technology (DHT) data for early detection of pregnancy-related conditions, using the comprehensive BUMP dataset to highlight the importance of individual variability and data heterogeneity. METHODS This BUMP study (Jan 2021 – May 2022) analyzed physiological and behavioral metrics, such as heart rate variability (HRV), sleep, and fatigue, in 256 individuals using Oura rings and self-reported surveys. Individual-level (N-of-1) trajectories were evaluated and compared with aggregate results to uncover personal and collective trends. A statistical method was developed to assess the influence of adverse events and severe symptoms, while case studies explored confounding and modifying factors underlying heterogeneity. Comprehensive statistical analysis included the coefficient of determination, Kolmogorov-Smirnov tests, likelihood ratio tests, and Welch’s t-tests, with inter-individual variability flagged based on high-variability thresholds. RESULTS Results revealed significant variability in HRV, sleep, and fatigue throughout pregnancy. For instance, only 4.76% of individuals had HRV inflection points at the aggregate week 33 inflection, with a 14.24% coefficient of variation. Our analysis found no significant p-values for demographic or pregnancy complication-based subgrouping, suggesting these factors alone do not drive the observed variability. Case studies further highlighted both intra- and inter-individual differences, emphasizing the importance of considering external factors like adverse events and severe symptoms. CONCLUSIONS Our findings show that aggregate wearable data often fails to generalize across populations, oversimplifying pregnancy-related physiological and subjective changes. This simplification can obscure individual trajectories, leading to generalized insights that may not reflect many pregnant women's experiences. Our results highlight the impact of heterogeneity on pregnancy outcomes, emphasizing the need to move beyond one-size-fits-all models and leverage DHT for personalized care.