CIMBID and Columbia Data Science Institute Launch New Biomedical Imaging Data Science Program
The Center for Innovation in Imaging Biomarkers and Integrated Diagnostics (CIMBID) has partnered with Columbia University's Data Science Institute (DSI) to launch the Biomedical Imaging & Data Science Scholars Program (DSI-BIDS), a new cross-disciplinary initiative designed to accelerate innovation at the intersection of artificial intelligence, biomedical imaging, and clinical translation.
The program pairs faculty-led projects with skilled data science students, with CIMBID and DSI providing financial and administrative support. Applications from faculty will be accepted twice each year, with the first deadline of August 5, 2026. Students apply for specific projects shortly after the faculty deadline.
DSI-BIDS aims to bring together faculty, trainees, and researchers from across Columbia to develop and apply advanced data science and AI methods to some of the most pressing challenges in medical imaging and precision medicine. By fostering collaboration across disciplines—including radiology, computer science, biomedical engineering, informatics, and public health—the DSI-BIDS Program aims to advance the discovery of imaging biomarkers, integrated diagnostics, and AI-driven tools that improve patient care.
As a founding partner of the initiative, CIMBID will play a central role in connecting expertise in biomedical imaging with Columbia's growing data science ecosystem. The program will provide opportunities for collaborative research, education, seminars, networking, and trainee engagement, helping to cultivate the next generation of leaders in imaging AI and data science.
The launch of the DSI-BIDS Program builds on CIMBID's mission to advance research, education, and clinical translation in AI-driven imaging biomarkers and integrated diagnostics through broad collaboration across Columbia University and its clinical and industry partners.
References
DSI-BIDS Selected Projects
The following projects were selected and matched with students for the Fall 2026 Data Science Institute Biomedical Imaging & Data Science Scholars Program (DSI-BIDS).
AI-Enabled Multimodal Neuroimaging Biomarkers of Brain Dynamics and Subcortical Structures in Depression and Stroke
- Principal Investigators: Sam Payabvash, MD, and Zhishun Wang, PhD
- Student Researcher: Andrew Tsai (MS in Data Science)
The project aims to develop more sensitive and individualized neuroimaging biomarkers for two major brain health conditions—depression and stroke—where current imaging measures (e.g., standard connectivity or regional volume) often fail to capture key differences in brain dynamics, disease heterogeneity, and treatment/recovery trajectories. The proposal targets two complementary gaps: (1) Functional brain dynamics in depression, using “brain criticality” theory (edge-of-chaos dynamics) measured from resting-state multi-echo fMRI via point-process/percolation approaches and phenomenological renormalization group (PRG) analyses. (2) Subcortical structural abnormalities in stroke (and reliability in normative data), moving beyond volume to localized surface-shape and geometry metrics (e.g., Gaussian curvature) derived from structural MRI, and relating them to behavioral/clinical outcomes (numeracy, aphasia severity). Ultimately, the project will combine dynamic, connectomic, morphometric, lesion, and clinical features into explainable AI models with uncertainty quantification to support individualized characterization and prediction.
Early Prediction of Neoadjuvant Chemotherapy Response Using Quantitative Harmonic Motion Imaging Biomarkers
- Principal Investigators: Katja Pinker-Domenig, MD, PhD, and Elisa Konofagou, PHD
- Student Researcher: Janavi Kolpekwar (MS in Data Science)
For breast cancer patients receiving neoadjuvant chemotherapy (NACT), the key desired outcome—pathologic complete response (pCR)—is only known at surgery, months after treatment starts. Existing early-response monitoring (e.g., serial MRI as in I-SPY2) is expensive, often requires IV contrast, and is not practical for frequent repeat imaging. Tumor shrinkage is also a late marker of response. This project addresses the need for early, low-cost, contrast-free ultrasound biomarkers by using Harmonic Motion Imaging (HMI) ultrasound elastography to quantify tumor stiffness changes (and spatial heterogeneity) between baseline and 3 weeks into NACT, then modeling these changes to predict pCR and Residual Cancer Burden (RCB) and potentially identify regions of drug resistance.
Incidence and Characterization of Lung Nodules in AATD
- Principal Investigators: Ritu Gill, MD, MPH, and Monica Goldklang, MD
- Student Researcher: Ruiling Li (MS in Data Science)
Alpha-1 antitrypsin deficiency (AATD) is a genetic cause of emphysema, often affecting patients with little or no smoking history. In a contemporary AATD cohort (Alpha-1 Biomarker Consortium, A1BC), incidental lung nodules have been reported in ~14% of participants, raising key unanswered questions: how often nodules occur and develop over time, whether specific emphysema phenotypes and smoke exposure are associated with higher lung cancer concern, and how to best triage incidental nodules in this rare-disease COPD population. The proposal addresses this by combining expert radiology reads with quantitative CT and AI tools (including Sybil and emphysema phenotyping) to build an AATD-specific risk prediction approach for nodule assessment.