Zhe He

Zekang Wang, Zhe He, Borong Zhang et al.

How should we analyze memory in deep RL? We introduce tools for analyzing policies under partial observability and revealing how agents use memory to make decisions. To utilize these tools, we present POPGym Arcade, a collection of Atari-inspired, hardware-accelerated environments sharing a single observation and action space. Each environment provides fully and partially observable variants, enabling counterfactual studies on observability. We find that controlled studies are necessary for fair comparisons and identify a pathology where value functions smear credit over irrelevant history. Using this pathology, we demonstrate how out-of-distribution scenarios can contaminate memory, perturbing the policy far into the future. Our code is available at https://github.com/bolt-research/popgym-arcade.

Gregor Stiglic, Leon Kopitar, Lucija Gosak et al.

Primary care professionals struggle to keep up to date with the latest scientific literature critical in guiding evidence-based practice related to their daily work. To help solve the above-mentioned problem, we employed generative artificial intelligence techniques based on large-scale language models to summarize abstracts of scientific papers. Our objective is to investigate the potential of generative artificial intelligence in diminishing the cognitive load experienced by practitioners, thus exploring its ability to alleviate mental effort and burden. The study participants were provided with two use cases related to preventive care and behavior change, simulating a search for new scientific literature. The study included 113 university students from Slovenia and the United States randomized into three distinct study groups. The first group was assigned to the full abstracts. The second group was assigned to the short abstracts generated by AI. The third group had the option to select a full abstract in addition to the AI-generated short summary. Each use case study included ten retrieved abstracts. Our research demonstrates that the use of generative AI for literature review is efficient and effective. The time needed to answer questions related to the content of abstracts was significantly lower in groups two and three compared to the first group using full abstracts. The results, however, also show significantly lower accuracy in extracted knowledge in cases where full abstract was not available. Such a disruptive technology could significantly reduce the time required for healthcare professionals to keep up with the most recent scientific literature; nevertheless, further developments are needed to help them comprehend the knowledge accurately.

Qizhang Feng, Jiayi Yuan, Forhan Bin Emdad et al.

Stroke is a significant cause of mortality and morbidity, necessitating early predictive strategies to minimize risks. Traditional methods for evaluating patients, such as Acute Physiology and Chronic Health Evaluation (APACHE II, IV) and Simplified Acute Physiology Score III (SAPS III), have limited accuracy and interpretability. This paper proposes a novel approach: an interpretable, attention-based transformer model for early stroke mortality prediction. This model seeks to address the limitations of previous predictive models, providing both interpretability (providing clear, understandable explanations of the model) and fidelity (giving a truthful explanation of the model's dynamics from input to output). Furthermore, the study explores and compares fidelity and interpretability scores using Shapley values and attention-based scores to improve model explainability. The research objectives include designing an interpretable attention-based transformer model, evaluating its performance compared to existing models, and providing feature importance derived from the model.

Neel Bhate, Ansh Mittal, Zhe He et al.

Demographics, Social determinants of health, and family history documented in the unstructured text within the electronic health records are increasingly being studied to understand how this information can be utilized with the structured data to improve healthcare outcomes. After the GPT models were released, many studies have applied GPT models to extract this information from the narrative clinical notes. Different from the existing work, our research focuses on investigating the zero-shot learning on extracting this information together by providing minimum information to the GPT model. We utilize de-identified real-world clinical notes annotated for demographics, various social determinants, and family history information. Given that the GPT model might provide text different from the text in the original data, we explore two sets of evaluation metrics, including the traditional NER evaluation metrics and semantic similarity evaluation metrics, to completely understand the performance. Our results show that the GPT-3.5 method achieved an average of 0.975 F1 on demographics extraction, 0.615 F1 on social determinants extraction, and 0.722 F1 on family history extraction. We believe these results can be further improved through model fine-tuning or few-shots learning. Through the case studies, we also identified the limitations of the GPT models, which need to be addressed in future research.

Weisi Liu, Zhe He, Xiaolei Huang

Time roots in applying language models for biomedical applications: models are trained on historical data and will be deployed for new or future data, which may vary from training data. While increasing biomedical tasks have employed state-of-the-art language models, there are very few studies have examined temporal effects on biomedical models when data usually shifts across development and deployment. This study fills the gap by statistically probing relations between language model performance and data shifts across three biomedical tasks. We deploy diverse metrics to evaluate model performance, distance methods to measure data drifts, and statistical methods to quantify temporal effects on biomedical language models. Our study shows that time matters for deploying biomedical language models, while the degree of performance degradation varies by biomedical tasks and statistical quantification approaches. We believe this study can establish a solid benchmark to evaluate and assess temporal effects on deploying biomedical language models.

Xiaoyu Wang, Haoyong Ouyang, Balu Bhasuran et al.

Accurate interpretation of lab results is crucial in clinical medicine, yet most patient portals use universal normal ranges, ignoring conditional factors like age and gender. This study introduces Lab-AI, an interactive system that offers personalized normal ranges using retrieval-augmented generation (RAG) from credible health sources. Lab-AI has two modules: factor retrieval and normal range retrieval. We tested these on 122 lab tests: 40 with conditional factors and 82 without. For tests with factors, normal ranges depend on patient-specific information. Our results show GPT-4-turbo with RAG achieved a 0.948 F1 score for factor retrieval and 0.995 accuracy for normal range retrieval. GPT-4-turbo with RAG outperformed the best non-RAG system by 33.5% in factor retrieval and showed 132% and 100% improvements in question-level and lab-level performance, respectively, for normal range retrieval. These findings highlight Lab-AI's potential to enhance patient understanding of lab results.

Chengwen Zhang, Chun Yu, Borong Zhuang et al.

Long-range Human-Robot Interaction (HRI) remains underexplored. Within it, Command Source Identification (CSI) - determining who issued a command - is especially challenging due to multi-user and distance-induced sensor ambiguity. We introduce HiSync, an optical-inertial fusion framework that treats hand motion as binding cues by aligning robot-mounted camera optical flow with hand-worn IMU signals. We first elicit a user-defined (N=12) gesture set and collect a multimodal command gesture dataset (N=38) in long-range multi-user HRI scenarios. Next, HiSync extracts frequency-domain hand motion features from both camera and IMU data, and a learned CSINet denoises IMU readings, temporally aligns modalities, and performs distance-aware multi-window fusion to compute cross-modal similarity of subtle, natural gestures, enabling robust CSI. In three-person scenes up to 34m, HiSync achieves 92.32% CSI accuracy, outperforming the prior SOTA by 48.44%. HiSync is also validated on real-robot deployment. By making CSI reliable and natural, HiSync provides a practical primitive and design guidance for public-space HRI.

Qiao Jin, Zhizheng Wang, Yifan Yang et al.

Clinical calculators play a vital role in healthcare by offering accurate evidence-based predictions for various purposes such as prognosis. Nevertheless, their widespread utilization is frequently hindered by usability challenges, poor dissemination, and restricted functionality. Augmenting large language models with extensive collections of clinical calculators presents an opportunity to overcome these obstacles and improve workflow efficiency, but the scalability of the manual curation process poses a significant challenge. In response, we introduce AgentMD, a novel language agent capable of curating and applying clinical calculators across various clinical contexts. Using the published literature, AgentMD has automatically curated a collection of 2,164 diverse clinical calculators with executable functions and structured documentation, collectively named RiskCalcs. Manual evaluations show that RiskCalcs tools achieve an accuracy of over 80% on three quality metrics. At inference time, AgentMD can automatically select and apply the relevant RiskCalcs tools given any patient description. On the newly established RiskQA benchmark, AgentMD significantly outperforms chain-of-thought prompting with GPT-4 (87.7% vs. 40.9% in accuracy). Additionally, we also applied AgentMD to real-world clinical notes for analyzing both population-level and risk-level patient characteristics. In summary, our study illustrates the utility of language agents augmented with clinical calculators for healthcare analytics and patient care.

Zhe He, Balu Bhasuran, Qiao Jin et al.

Lab results are often confusing and hard to understand. Large language models (LLMs) such as ChatGPT have opened a promising avenue for patients to get their questions answered. We aim to assess the feasibility of using LLMs to generate relevant, accurate, helpful, and unharmful responses to lab test-related questions asked by patients and to identify potential issues that can be mitigated with augmentation approaches. We first collected lab test results related question and answer data from Yahoo! Answers and selected 53 QA pairs for this study. Using the LangChain framework and ChatGPT web portal, we generated responses to the 53 questions from four LLMs including GPT-4, Meta LLaMA 2, MedAlpaca, and ORCA_mini. We first assessed the similarity of their answers using standard QA similarity-based evaluation metrics including ROUGE, BLEU, METEOR, BERTScore. We also utilized an LLM-based evaluator to judge whether a target model has higher quality in terms of relevance, correctness, helpfulness, and safety than the baseline model. Finally, we performed a manual evaluation with medical experts for all the responses to seven selected questions on the same four aspects. The results of Win Rate and medical expert evaluation both showed that GPT-4's responses achieved better scores than all the other LLM responses and human responses on all four aspects (relevance, correctness, helpfulness, and safety). However, LLM responses occasionally also suffer from a lack of interpretation in one's medical context, incorrect statements, and lack of references. We find that compared to other three LLMs and human answer from the Q&A website, GPT-4's responses are more accurate, helpful, relevant, and safer. However, there are cases which GPT-4 responses are inaccurate and not individualized. We identified a number of ways to improve the quality of LLM responses.

Qiao Jin, Nicholas Wan, Robert Leaman et al.

Large language models (LLMs) represent a transformative class of AI tools capable of revolutionizing various aspects of healthcare by generating human-like responses across diverse contexts and adapting to novel tasks following human instructions. Their potential application spans a broad range of medical tasks, such as clinical documentation, matching patients to clinical trials, and answering medical questions. In this primer paper, we propose an actionable guideline to help healthcare professionals more efficiently utilize LLMs in their work, along with a set of best practices. This approach consists of several main phases, including formulating the task, choosing LLMs, prompt engineering, fine-tuning, and deployment. We start with the discussion of critical considerations in identifying healthcare tasks that align with the core capabilities of LLMs and selecting models based on the selected task and data, performance requirements, and model interface. We then review the strategies, such as prompt engineering and fine-tuning, to adapt standard LLMs to specialized medical tasks. Deployment considerations, including regulatory compliance, ethical guidelines, and continuous monitoring for fairness and bias, are also discussed. By providing a structured step-by-step methodology, this tutorial aims to equip healthcare professionals with the tools necessary to effectively integrate LLMs into clinical practice, ensuring that these powerful technologies are applied in a safe, reliable, and impactful manner.

James V. Roggeveen, Erik Y. Wang, Will Flintoft et al.

Large language models (LLMs) have shown remarkable progress in mathematical problem-solving, but evaluation has largely focused on problems that have exact analytical solutions or involve formal proofs, often overlooking approximation-based problems ubiquitous in applied science and engineering. To fill this gap, we build on prior work and present HARDMath2, a dataset of 211 original problems covering the core topics in an introductory graduate applied math class, including boundary-layer analysis, WKB methods, asymptotic solutions of nonlinear partial differential equations, and the asymptotics of oscillatory integrals. This dataset was designed and verified by the students and instructors of a core graduate applied mathematics course at Harvard. We build the dataset through a novel collaborative environment that challenges students to write and refine difficult problems consistent with the class syllabus, peer-validate solutions, test different models, and automatically check LLM-generated solutions against their own answers and numerical ground truths. Evaluation results show that leading frontier models still struggle with many of the problems in the dataset, highlighting a gap in the mathematical reasoning skills of current LLMs. Importantly, students identified strategies to create increasingly difficult problems by interacting with the models and exploiting common failure modes. This back-and-forth with the models not only resulted in a richer and more challenging benchmark but also led to qualitative improvements in the students' understanding of the course material, which is increasingly important as we enter an age where state-of-the-art language models can solve many challenging problems across a wide domain of fields.

Guanchen Wu, Yuzhang Xie, Huanwei Wu et al.

Integrating novel medical concepts and relationships into existing ontologies can significantly enhance their coverage and utility for both biomedical research and clinical applications. Clinical notes, as unstructured documents rich with detailed patient observations, offer valuable context-specific insights and represent a promising yet underutilized source for ontology extension. Despite this potential, directly leveraging clinical notes for ontology extension remains largely unexplored. To address this gap, we propose CLOZE, a novel framework that uses large language models (LLMs) to automatically extract medical entities from clinical notes and integrate them into hierarchical medical ontologies. By capitalizing on the strong language understanding and extensive biomedical knowledge of pre-trained LLMs, CLOZE effectively identifies disease-related concepts and captures complex hierarchical relationships. The zero-shot framework requires no additional training or labeled data, making it a cost-efficient solution. Furthermore, CLOZE ensures patient privacy through automated removal of protected health information (PHI). Experimental results demonstrate that CLOZE provides an accurate, scalable, and privacy-preserving ontology extension framework, with strong potential to support a wide range of downstream applications in biomedical research and clinical informatics.

Balu Bhasuran, Mattia Prosperi, Karim Hanna et al.

This study evaluates causal reasoning in large language models (LLMs) using 99 clinically grounded laboratory test scenarios aligned with Pearl's Ladder of Causation: association, intervention, and counterfactual reasoning. We examined common laboratory tests such as hemoglobin A1c, creatinine, and vitamin D, and paired them with relevant causal factors including age, gender, obesity, and smoking. Two LLMs - GPT-o1 and Llama-3.2-8b-instruct - were tested, with responses evaluated by four medically trained human experts. GPT-o1 demonstrated stronger discriminative performance (AUROC overall = 0.80 +/- 0.12) compared to Llama-3.2-8b-instruct (0.73 +/- 0.15), with higher scores across association (0.75 vs 0.72), intervention (0.84 vs 0.70), and counterfactual reasoning (0.84 vs 0.69). Sensitivity (0.90 vs 0.84) and specificity (0.93 vs 0.80) were also greater for GPT-o1, with reasoning ratings showing similar trends. Both models performed best on intervention questions and worst on counterfactuals, particularly in altered outcome scenarios. These findings suggest GPT-o1 provides more consistent causal reasoning, but refinement is required before adoption in high-stakes clinical applications.

Hanshu Rao, Weisi Liu, Haohan Wang et al.

Synthetic data generation--mitigating data scarcity, privacy concerns, and data quality challenges in biomedical fields--has been facilitated by rapid advances of large language models (LLMs). This scoping review follows PRISMA-ScR guidelines and synthesizes 59 studies, published between 2020 and 2025 and collected from PubMed, ACM, Web of Science, and Google Scholar. The review systematically examines biomedical research and application trends in synthetic data generation, emphasizing clinical applications, methodologies, and evaluations. Our analysis identifies data modalities of unstructured texts (78.0%), tabular data (13.6%), and multimodal sources (8.4%); generation methods of prompting (72.9%), fine-tuning (22.0%) LLMs and specialized model (5.1%); and heterogeneous evaluations of intrinsic metrics (27.1%), human-in-the-loop assessments (55.9%), and LLM-based evaluations (13.6%). The analysis addresses current limitations in what, where, and how health professionals can leverage synthetic data generation for biomedical domains. Our review also highlights challenges in adaption across clinical domains, resource and model accessibility, and evaluation standardizations.

Zekang Wang, Zhe He, Borong Zhang et al.

How should we analyze memory in deep RL? We introduce mathematical tools for fairly analyzing policies under partial observability and revealing how agents use memory to make decisions. To utilize these tools, we present POPGym Arcade, a collection of Atari-inspired, hardware-accelerated, pixel-based environments sharing a single observation and action space. Each environment provides fully and partially observable variants, enabling counterfactual studies on observability. We find that controlled studies are necessary for fair comparisons, and identify a pathology where value functions smear credit over irrelevant history. With this pathology, we demonstrate how out-of-distribution scenarios can contaminate memory, perturbing the policy far into the future, with implications for sim-to-real transfer and offline RL.

Balu Bhasuran, Qiao Jin, Yuzhang Xie et al.

Differential diagnosis is crucial for medicine as it helps healthcare providers systematically distinguish between conditions that share similar symptoms. This study assesses the impact of lab test results on differential diagnoses (DDx) made by large language models (LLMs). Clinical vignettes from 50 case reports from PubMed Central were created incorporating patient demographics, symptoms, and lab results. Five LLMs GPT-4, GPT-3.5, Llama-2-70b, Claude-2, and Mixtral-8x7B were tested to generate Top 10, Top 5, and Top 1 DDx with and without lab data. A comprehensive evaluation involving GPT-4, a knowledge graph, and clinicians was conducted. GPT-4 performed best, achieving 55% accuracy for Top 1 diagnoses and 60% for Top 10 with lab data, with lenient accuracy up to 80%. Lab results significantly improved accuracy, with GPT-4 and Mixtral excelling, though exact match rates were low. Lab tests, including liver function, metabolic/toxicology panels, and serology/immune tests, were generally interpreted correctly by LLMs for differential diagnosis.

Zhe He, Shubo Tian, Arslan Erdengasileng et al.

Subtyping of Alzheimer's disease (AD) can facilitate diagnosis, treatment, prognosis and disease management. It can also support the testing of new prevention and treatment strategies through clinical trials. In this study, we employed spectral clustering to cluster 29,922 AD patients in the OneFlorida Data Trust using their longitudinal EHR data of diagnosis and conditions into four subtypes. These subtypes exhibit different patterns of progression of other conditions prior to the first AD diagnosis. In addition, according to the results of various statistical tests, these subtypes are also significantly different with respect to demographics, mortality, and prescription medications after the AD diagnosis. This study could potentially facilitate early detection and personalized treatment of AD as well as data-driven generalizability assessment of clinical trials for AD.

Francois Modave, Yunpeng Zhao, Janice Krieger et al.

Among American women, the rate of breast cancer is only second to lung cancer. An estimated 12.4% women will develop breast cancer over the course of their lifetime. The widespread use of social media across the socio-economic spectrum offers unparalleled ways to facilitate information sharing, in particular as it pertains to health. Social media is also used by many healthcare stakeholders, ranging from government agencies to healthcare industry, to disseminate health information and to engage patients. The purpose of this study is to investigate people's perceptions and attitudes relate to breast cancer, especially those that are related to physical activities, on Twitter. To achieve this, we first identified and collected tweets related to breast cancer; and then used topic modeling and sentiment analysis techniques to understanding discussion themes and quantify Twitter users' perceptions and emotions w.r.t breast cancer to answer 5 research questions.

Shaeke Salman, Seyedeh Neelufar Payrovnaziri, Xiuwen Liu et al.

Deep neural networks have achieved remarkable success in various challenging tasks. However, the black-box nature of such networks is not acceptable to critical applications, such as healthcare. In particular, the existence of adversarial examples and their overgeneralization to irrelevant, out-of-distribution inputs with high confidence makes it difficult, if not impossible, to explain decisions by such networks. In this paper, we analyze the underlying mechanism of generalization of deep neural networks and propose an ($n$, $k$) consensus algorithm which is insensitive to adversarial examples and can reliably reject out-of-distribution samples. Furthermore, the consensus algorithm is able to improve classification accuracy by using multiple trained deep neural networks. To handle the complexity of deep neural networks, we cluster linear approximations of individual models and identify highly correlated clusters among different models to capture feature importance robustly, resulting in improved interpretability. Motivated by the importance of building accurate and interpretable prediction models for healthcare, our experimental results on an ICU dataset show the effectiveness of our algorithm in enhancing both the prediction accuracy and the interpretability of deep neural network models on one-year patient mortality prediction. In particular, while the proposed method maintains similar interpretability as conventional shallow models such as logistic regression, it improves the prediction accuracy significantly.

Seyedeh Neelufar Payrovnaziri, Laura A. Barrett, Daniel Bis et al.

Predicting the risk of mortality for patients with acute myocardial infarction (AMI) using electronic health records (EHRs) data can help identify risky patients who might need more tailored care. In our previous work, we built computational models to predict one-year mortality of patients admitted to an intensive care unit (ICU) with AMI or post myocardial infarction syndrome. Our prior work only used the structured clinical data from MIMIC-III, a publicly available ICU clinical database. In this study, we enhanced our work by adding the word embedding features from free-text discharge summaries. Using a richer set of features resulted in significant improvement in the performance of our deep learning models. The average accuracy of our deep learning models was 92.89% and the average F-measure was 0.928. We further reported the impact of different combinations of features extracted from structured and/or unstructured data on the performance of the deep learning models.

Zhe He, Adrian Spurr, Xucong Zhang et al.

Gaze redirection is the task of changing the gaze to a desired direction for a given monocular eye patch image. Many applications such as videoconferencing, films, games, and generation of training data for gaze estimation require redirecting the gaze, without distorting the appearance of the area surrounding the eye and while producing photo-realistic images. Existing methods lack the ability to generate perceptually plausible images. In this work, we present a novel method to alleviate this problem by leveraging generative adversarial training to synthesize an eye image conditioned on a target gaze direction. Our method ensures perceptual similarity and consistency of synthesized images to the real images. Furthermore, a gaze estimation loss is used to control the gaze direction accurately. To attain high-quality images, we incorporate perceptual and cycle consistency losses into our architecture. In extensive evaluations we show that the proposed method outperforms state-of-the-art approaches in terms of both image quality and redirection precision. Finally, we show that generated images can bring significant improvement for the gaze estimation task if used to augment real training data.

Min Zhao, Jie Chen, Zhe He

Spectral unmixing is an important and challenging problem in hyperspectral data processing. This topic has been extensively studied and a variety of unmixing algorithms have been proposed in the literature. However, the lack of publicly available dataset with ground-truth makes it difficult to evaluate and compare the performance of unmixing algorithms in a quantitative and objective manner. Most of the existing works rely on the use of numerical synthetic data and an intuitive inspection of the results of real data. To alleviate this dilemma, in this study, we design several experimental scenes in our laboratory, including printed checkerboards, mixed quartz sands, and reflection with a vertical board. A dataset is then created by imaging these scenes with the hyperspectral camera in our laboratory, providing 36 mixtures with more than 130, 000 pixels with 256 wavelength bands ranging from 400nm to 1000nm. The experimental settings are strictly controlled so that pure material spectral signatures and material compositions are known. To the best of our knowledge, this dataset is the first publicly available dataset created in a systematic manner with ground-truth for spectral unmixing. Some typical linear and nonlinear unmixing algorithms are also tested with this dataset and lead to meaningful results.

Laura A. Barrett, Seyedeh Neelufar Payrovnaziri, Jiang Bian et al.

Heart disease remains the leading cause of death in the United States. Compared with risk assessment guidelines that require manual calculation of scores, machine learning-based prediction for disease outcomes such as mortality can be utilized to save time and improve prediction accuracy. This study built and evaluated various machine learning models to predict one-year mortality in patients diagnosed with acute myocardial infarction or post myocardial infarction syndrome in the MIMIC-III database. The results of the best performing shallow prediction models were compared to a deep feedforward neural network (Deep FNN) with back propagation. We included a cohort of 5436 admissions. Six datasets were developed and compared. The models applying Logistic Model Trees (LMT) and Simple Logistic algorithms to the combined dataset resulted in the highest prediction accuracy at 85.12% and the highest AUC at .901. In addition, other factors were observed to have an impact on outcomes as well.