About

Daniel Rubin is a Professor of Biomedical Data Science and of Radiology at Stanford University, with additional appointments in Integrative Biomedical Imaging Informatics, Medicine (Biomedical Informatics Research), Ophthalmology, and Computer Science. He is affiliated with the Center for Artificial Intelligence in Medicine & Imaging (AIMI) at Stanford. His research focuses on the application of artificial intelligence and data science to medicine and imaging, aiming to advance healthcare through innovative computational methods. Rubin's work involves integrating biomedical data, developing imaging informatics, and contributing to the field of AI for healthcare, leveraging his multidisciplinary expertise to improve diagnostic and treatment processes.

Research topics

• Computer Science
• Artificial Intelligence
• Medicine
• Machine Learning
• Radiology
• Data Mining
• Internal medicine
• Medical physics
• Nuclear medicine
• Physics
• Pathology
• Genetics
• Ophthalmology
• Systems engineering
• Biology
• Risk analysis (engineering)
• Data science
• Algorithm
• Engineering
• Oncology

Selected publications

• Extraction of distant recurrence sites for breast cancer patients from free-text clinical notes using large language models

  Journal of Biomedical Informatics · 2026-03-29

  articleOpen access
  PublisherOA PDFDOI

• Out-of-the-Box Large Language Models for Detecting and Classifying Critical Findings in Radiology Reports Using Various Prompt Strategies

  American Journal of Roentgenology · 2025-09-10 · 3 citations

  articleSenior author

  . The study shows a role of contemporary general-purpose models in adapting to specialized medical tasks using minimal data annotation.

  PublisherDOI

• Creation of Radiology Teaching Content with STELLA—A STandardized Electronic Learning Library and Application Platform

  Academic Radiology · 2025-12-26

  articleSenior author
  PublisherDOI

• Abstract P2-04-17: The MIRACCL Portal for Comparing Patient and PDX Response Using Cancer Image Features and Genomics in Co-Clinical Breast Cancer Trials

  Clinical Cancer Research · 2025-06-13

  article

  Abstract Introduction: The Molecular and Imaging Response Analysis of Co-Clinical Trials (MIRACCL, https://miraccl.research.bcm.edu/ ) platform was developed beginning three years ago to support the co-clinical breast cancer RESPONSE trial at Baylor College of Medicine. The goal of MIRACCL is to enable parallel studies which apply the same protocol to patients in a clinical trial and patient-derived xenograft (PDX) cohorts. MIRACCL evaluates treatment response by electronically measuring the changes in serial MRI. The response assessments include the measurement of multiple imagine features such as apparent diffusion coefficient (ADC) and signal enhancement ratio (SER) at pre-treatment and post-treatment along with samples taken for assessment of the genomic changes.Methods: To achieve this goal, we employee the use of web-technologies to enable visual and quantitative comparison of the imaging and genomics. Investigators at Baylor College of Medicine leveraged their expertise in patient derived xenograft development and model study implementation to generate the PDX imaging dataset and collect the model samples for sequencing. The University of Texas at Austin provided centralized patient and PDX image normalization, segmentation, and analysis while the Biomedical Informatics team at Stanford University provided the image visualization tool and imaging response assessment. Due to the difference in data modalities, data sources, and temporal collection of data it was necessary to place the data within the context of the study design and provide visual and quantitative comparisons of the study outcomes for various time points. The MIRACCL features created to achieve this goal include tabular cohort annotations, a side-by-side visualization of imaging response distributions, and a comparison of the upregulated and down regulated gene expression between cohorts. One of the imaging methods deployed in MIRACCL to assess treatment response measures response by tumor longest diameter which is then categorized by RECIST. Additional imaging comparison methods include signal enhancement ratio (SER), apparent diffusion coefficient (ADC), and tumor volume. Samples for sequencing taken at pre-treatment and throughout the design of the study are used to identify gene expression changes brought about by treatment. In the Omics module of MIRACCL, the 500 most frequently upregulated and 500 down regulated genes for each cohort are displayed based on user selected time points and imaging feature of interest. A Venn diagram is generated to identify the genes which are commonly regulated in both cohorts.Results & Conclusion: While these features have provided multiple methods of comparing the outcome of co-clinical trials, the research team desired to know the significance of these correlations and differences between the two cohorts. Consequently, the Analytic module was implemented in MIRACCL this past year. The analytics module focuses on two hypotheses: 1) PDX models of similar subtyping will respond in a similar manner to the patients enrolled in the trial and 2) Response can be predicted while on-treatment to determine if changes to treatment are warranted. The change in tumor quantifications from on-treatment to baseline were statistically correlated to changes in tumor quantifications from post-treatment to baseline to determine if response could be determined while currently on-treatment. The p value and r value of the spearman correlation are provided to determine the significance and clustering of the cohort. The enhancements afforded by MIRACCL’s Analytics module summarize the treatment response of the patient and PDX cohorts and effectively address the hypothesis of the REPONSE trial. MIRACCL is now available for expansion as a tool for other trials co-clinical trials. Citation Format: Heidi Dowst, Fei Zheng, Emel Alkim, Apollo McOwiti, Ram Rajaram Srinivasan, David Hormuth, Thomas Yankeelov, Daniel Rubin, Michael T. Lewis. The MIRACCL Portal for Comparing Patient and PDX Response Using Cancer Image Features and Genomics in Co-Clinical Breast Cancer Trials [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr P2-04-17.

  PublisherDOI

• Open-Source Hybrid Large Language Model Integrated System for Extraction of Breast Cancer Treatment Pathway From Free-Text Clinical Notes

  JCO Clinical Cancer Informatics · 2025-06-01 · 3 citations

  article

  PURPOSE: Automated curation of breast cancer treatment data with minimal human involvement could accelerate the collection of statewide and nationwide evidence for patient management and assessing the effectiveness of treatment pathways. The primary challenges are the complexity and inconsistency of structured clinical data streams and accurate extraction of this information from free-text clinical narratives. MATERIALS AND METHODS: We proposed a hybrid two-phase information extraction framework that combined a Unified Medical Language System parser (phase-1) with a fine-tuned large language model (LLM; phase-2) to extract longitudinal treatment timelines from time-stamped clinical notes. Our framework was developed through end-to-end joint learning as a question-answering model, where the model was trained to simultaneously answer five questions, each corresponding to a specific treatment. RESULTS: We fine-tuned and internally validated the model on 26,692 patients with breast cancer (diagnosed between 2013 and 2020) receiving treatment at Mayo Clinic and externally validated the model on 162 randomly selected patients from Stanford Healthcare. Zero-shot LLM (out-of-the-box) had high specificity but low sensitivity, indicating that although these frameworks are useful for generic language understanding, they are lacking in terms of targeted clinical tasks. The proposed model achieved 0.942 average AUROC on the internal and 0.924 on the external data, demonstrating only marginal drop in performance when evaluated on external. The proposed model also achieved better trade-off between sensitivity (average: 79.2%) and specificity (average: 76.2%) compared with rule-based (average sensitivity: 70.5%, average specificity: 68.1%) and structured codes (average sensitivity: 64.1%, average specificity: 83.5%). CONCLUSION: The proposed framework can extract temporal information about cancer treatments from various time-stamped clinic notes, regardless of the setting of treatment administration (inpatient or outpatient) or time frame. To support the cancer research community for such data curation and longitudinal analysis, we have packaged the code as a docker image, which needs minimal system reconfiguration and shared with an open-source academic license.

  PublisherDOI

• Automated Extraction of Patient-Centered Outcomes After Breast Cancer Treatment: An Open-Source Large Language Model–Based Toolkit

  JCO Clinical Cancer Informatics · 2024-08-01 · 9 citations

  articleOpen access

  PURPOSE: Patient-centered outcomes (PCOs) are pivotal in cancer treatment, as they directly reflect patients' quality of life. Although multiple studies suggest that factors affecting breast cancer-related morbidity and survival are influenced by treatment side effects and adherence to long-term treatment, such data are generally only available on a smaller scale or from a single center. The primary challenge with collecting these data is that the outcomes are captured as free text in clinical narratives written by clinicians. MATERIALS AND METHODS: Given the complexity of PCO documentation in these narratives, computerized methods are necessary to unlock the wealth of information buried in unstructured text notes that often document PCOs. Inspired by the success of large language models (LLMs), we examined the adaptability of three LLMs, GPT-2, BioGPT, and PMC-LLaMA, on PCO tasks across three institutions, Mayo Clinic, Emory University Hospital, and Stanford University. We developed an open-source framework for fine-tuning LLM that can directly extract the five different categories of PCO from the clinic notes. RESULTS: We found that these LLMs without fine-tuning (zero-shot) struggle with challenging PCO extraction tasks, displaying almost random performance, even with some task-specific examples (few-shot learning). The performance of our fine-tuned, task-specific models is notably superior compared with their non-fine-tuned LLM models. Moreover, the fine-tuned GPT-2 model has demonstrated a significantly better performance than the other two larger LLMs. CONCLUSION: Our discovery indicates that although LLMs serve as effective general-purpose models for tasks across various domains, they require fine-tuning when applied to the clinician domain. Our proposed approach has the potential to lead more efficient, adaptable models for PCO information extraction, reducing reliance on extensive computational resources while still delivering superior performance for specific tasks.

  PublisherOA PDFDOI

• Addressing Catastrophic Forgetting by Modulating Global Batch Normalization Statistics for Medical Domain Expansion

  Lecture notes in computer science · 2024-10-02

  book-chapter
  PublisherDOI

• Mirrored X-Net: Joint classification and contrastive learning for weakly supervised GA segmentation in SD-OCT

  Pattern Recognition · 2024-04-18 · 7 citations

  article
  PublisherDOI

• Privacy preservation for federated learning in health care

  Patterns · 2024-07-01 · 135 citations

  reviewOpen access

  Artificial intelligence (AI) shows potential to improve health care by leveraging data to build models that can inform clinical workflows. However, access to large quantities of diverse data is needed to develop robust generalizable models. Data sharing across institutions is not always feasible due to legal, security, and privacy concerns. Federated learning (FL) allows for multi-institutional training of AI models, obviating data sharing, albeit with different security and privacy concerns. Specifically, insights exchanged during FL can leak information about institutional data. In addition, FL can introduce issues when there is limited trust among the entities performing the compute. With the growing adoption of FL in health care, it is imperative to elucidate the potential risks. We thus summarize privacy-preserving FL literature in this work with special regard to health care. We draw attention to threats and review mitigation approaches. We anticipate this review to become a health-care researcher's guide to security and privacy in FL.

  PublisherOA PDFDOI

• Assessing breast cancer chemotherapy response in radiology and pathology reports via a large language model

  2024-04-02

  article

  A wealth of medical knowledge is used to make clinical decisions, yet treatment or disease outcomes are challenging to assess without clinical trials. However, clinical trials take time, are expensive, and are impossible to perform for every decision. One approach to systematically assess treatment outcomes involves the retrospective analysis of clinical notes, e.g., radiology and pathology reports. While such studies are often performed by clinicians who manually review the notes and other information, such retrospective analysis can benefit from the automated parsing of radiology and pathology reports to provide systematic framework to extract outcome information. In this study, we used a large language model, i.e., ChatGPT (GPT-3.5), to parse 267 radiology and pathology reports and extract information related to response to neoadjuvant chemotherapy in patients with breast cancer. Our study includes a heterogeneous group of 89 women who underwent neoadjuvant therapy and underwent two MRI exams, pre- and post-therapy, followed by surgery (lumpectomy or mastectomy). We assessed the treatment response based on clinical reports from the post-therapy surgical excision. From the reports, we extracted the number of lesions, their anatomic location, and size. Our study provides insight into neoadjuvant chemotherapy response, indicating that even cases with complete MRI response can still have residual invasive breast carcinoma (1/3 of subjects), and, on the other hand, even those with reduced MRI response (⪅30% reduction in tumor size) can have no residual tumor at surgery, indicating that when cancer responds to treatment, it may not be captured by the MRI. The large language model achieved sensitivities of 84-94% in extracting the information from radiology reports, but had lower performance in the pathology reports, 72-87%, where more information is provided in free format. While this study is preliminary and performed in a small cohort, it illustrates the complexity of outcome prediction using radiology images.

  PublisherDOI

Recent grants

• Qualification and Deployment of Imaging Biomarkers of Cancer Treatment Response

  NIH · $3.1M · 2015–2022

• Distributed Learning of Deep Learning Models for Cancer Research

  NIH · $1.2M · 2019–2022

• NIH Grant U01CA142555

  NIH · $2.9M · 2017

• Cancer Center Administration

  NIH · $34.9M · 2007–2027

• NIH Grant U01CA187947

  NIH · $2.9M · 2020

Frequent coauthors

• Sandy Napel

  Stanford University

  63 shared

• Imon Banerjee

  57 shared

• Curtis P. Langlotz

  Palo Alto University

  52 shared

• Theodore Leng

  Stanford University

  50 shared

• Luís de Sisternes

  Carl Zeiss (United States)

  45 shared

• Daniel J. Vreeman

  41 shared

• Kenneth C. Wang

  Baltimore VA Medical Center

  40 shared

• Eliot L. Siegel

  University of Maryland, Baltimore

  40 shared