The DSI Health Analytics Center works to improve the health of individuals and the health care system through data-driven methods and understanding of health processes.
Our work builds upon the work of teams of Columbia researchers in medicine, biology, public health, informatics, computer science, applied mathematics, and statistics. The Health Analytics Center is located at the Columbia University Medical Center. Learn More.
Details & Recording
Thursday, May 4, 2023 (10:00 AM – 2:00 PM ET) – Hybrid
Location: Northwest Corner Building, 14th Floor (DSI Suite) – 550 W 120th St, New York, NY 10027
Speakers & Agenda
Agenda for May 4, 2023 Each speaker to give a 20 minute talk followed by 5 minutes of Q&A
10:00 AM: Coffee & Welcome
Introductions from DSI Health Analytics Center Chairs
10:10 AM: Talk 1
Elham Azizi, Assistant Professor of Biomedical Engineering and Herbert and Florence Irving Assistant Professor of Cancer Data Research (in the Herbert and Florence Irving Institute for Cancer Dynamics and in the Herbert Irving Comprehensive Cancer Center)
Probabilistic Modeling of Dynamics of the Tumor Microenvironment
10:35 AM: Talk 2
Carri Chan, Professor of Business in the Division of Decision, Risk, and Operations and the Faculty Director of the Healthcare and Pharmaceutical Management Program, Columbia Business School
Prediction-Driven Surge Planning with Application in the Emergency Department
11:00 AM: Talk 3
Sachin Jambawalikar, Chief Medical Physicist in Department of Radiology at Columbia University Irving Medical Center and NewYork-Presbyterian
Enhancing Radiology through Explainable AI
11:25 AM: Break (10 min)
11:35 AM: Talk 4
Paul Sajda, Vikram S. Pandit Professor of Biomedical Engineering and Professor of Radiology (Physics) and Electrical Engineering, Columbia Engineering
Closed-loop Non-invasive Neural Therapeutics
12:00 PM: Talk 5
Christine Hendon, Associate Professor of Electrical Engineering, Columbia Engineering
Monitoring and Guidance of Cardiac Ablation Therapy with Optics
12:25 PM: Talk 6
Frank Provenzano, Assistant Professor of Neurological Sciences (in Neurology in The Taub Institute), Columbia University
Elham Azizi: Probabilistic Modeling of Dynamics of the Tumor Microenvironment
Cancer therapies succeed only in a subset of patients partly due to the heterogeneity of cells across and within Cancer therapies succeed only in a subset of patients partly due to the heterogeneity of cells across and within tumors. Single-cell and spatial genomic technologies present exciting opportunities to characterize unknown cell types in complex tissues such as tumor microenvironments and elucidate their interactions, circuitry, and role in driving response to therapies. However, analyzing and integrating single-cell data across conditions, patients, time points, and data modalities involve significant statistical and computational challenges. I will present a set of probabilistic and deep generative models developed for addressing these problems and modeling temporal and spatial dynamics of key immune subsets defining cancer progression and response to immunotherapy. I will also present Starfysh for spatial mapping of heterogeneous cell states and crosstalk in complex tissues, from the integration of spatial transcriptomics and histology images.
Carri Chan: Prediction-Driven Surge Planning with Application in the Emergency Department
Determining emergency department (ED) nurse staffing decisions to balance the quality of service and staffing cost can be extremely challenging, especially when there is a high level of uncertainty in patient-demand. Increasing data availability and continuing advancements in predictive analytics provide an opportunity to mitigate demand uncertainty by utilizing demand forecasts. In this work, we study a two-stage prediction- driven staffing framework where the prediction models are integrated with the base (made weeks in advance) and surge (made nearly real-time) staffing decisions in the ED. We quantify the benefit of having the ability to use the more expensive surge staffing and identify the importance of balancing demand uncertainty versus demand stochasticity. We also propose a near-optimal two-stage staffing policy that is straightforward to interpret and implement. Lastly, we develop a unified framework that combines parameter estimation, real- time demand forecasts, and capacity sizing in the ED. High-fidelity simulation experiments for the ED demonstrate that the proposed framework can reduce annual staffing costs by 11%–16% ($2 M–$3 M) while guaranteeing timely access to care.
Sachin Jambawalikar: Enhancing Radiology through Explainable AI
Radiological imaging has become an essential tool in modern medicine, playing a critical role in diagnosing and treating a wide range of medical conditions, from broken bones to cancer. However, with the advancements in medical imaging technology, its demand has increased, adding pressure on radiologists to provide efficient and high-quality services despite limited resources and personnel. The shortage of qualified personnel can impact diagnoses and result in staff burnout. Additionally, advances in imaging technology generate significant amounts of data, leading to delays and errors in reporting. Numerous artificial intelligence (AI) and machine learning (ML) algorithms have been developed and introduced in recent years to improve diagnostic accuracy and efficiency. One of the ongoing challenges in Radiology AI is to develop systems that can explain their reasoning or allow humans to interpret their output. Traditionally, AI algorithms have been used as black boxes and are vulnerable to simple attacks, raising concerns about their ability to be trusted. These limitations are especially a problem in medical settings where the decision-making process needs to be transparent, interpretable, and trustworthy. Explainable AI (XAI) is a new approach to machine learning that aims to address these limitations, making it a promising tool for enhancing radiology. This talk will explore the potential benefits of XAI in radiology, such as increased transparency, accountability, and trust in the decision-making process thereby enhancing accuracy, efficiency, and quality of care provided to patients.
Paul Sajda: Closed-loop Non-invasive Neural Therapeutics
Precision and personalized delivery of neurostimulation, such as transcranial magnetic stimulation (TMS), is believed to be critical for the effective treatment of psychiatric diseases such as major depressive disorder (MDD). In this talk, I will present our work that uses an integrated instrument consisting of simultaneous fMRI, EEG, and TMS (which we term “fET”) to determine the precise timing of TMS stimulation to deliver to an individual to maximize engagement of the therapeutic target—in this case, the anterior cingulate cortex. I will present results from a double-blind clinical trial that uses the fET instrument to set individualized parameters of a closed-loop EEG-TMS neurostimulator for treating MDD. We demonstrate that this closed-loop precision delivery of stimulation, which is matched to specific brain dynamics, affects brain dynamics, functional and effective connectivity, and ultimately clinical outcome of the treatment. Our approach enables new types of non-invasive neural therapeutics which are personalized to an individual as well as a disease or condition.
Christine P Hendon: Monitoring and Guidance of Cardiac Ablation Therapy with Optics
The research goals of the Structure-Function Imaging Laboratory are to develop platform optical imaging systems to enable structure-function analysis of biological organ systems. Towards this goal, we develop optical coherence tomography (OCT) and near infrared spectroscopy (NIRS) systems and automated processing tools to correlate tissue microstructure to electrical conduction and mechanical contraction. Within this talk, I will highlight our effort towards monitoring and guidance of radiofrequency ablation therapy with optical technologies. There is typically an underlying substrate due to remodeling with the development of scar tissue or fibrosis that is the cause of the abnormalities in conduction patterns. Therefore, better understanding how the microstructure of the myocardium influences electrical conduction will greatly inform these therapeutic procedures. We propose to use optical imaging and spectroscopy as a means to monitor and guide radiofrequency ablation treatment of cardiac arrhythmias, which will directly interrogate the tissue for characterization in real time. I will present analysis of OCT and NIRS optical signals, metrics to extract information on energy delivery, remodeling and composition, and fiber orientations within human hearts. We have shown that NIRS integrated with an RFA catheter is capable of extracting tissue information as deep as 4mm. NIRS can also assess tissue-catheter contact with high accuracy and discriminate ablated from unablated tissue. Importantly, gaps can be readily identified and maps of lesion thickness calculated. OCT detects distinguishing features between regions of myocardium within the left atrium and regions of only venous endothelium, media, and adventitia – that is, regions of transmural connective tissues – inside the pulmonary veins. Detailed tissue characteristics such as endocardial thickness, myointimal thickening, and fibrosis could also be determined. Together, this will provide the foundation for optical imaging guidance providing information on tissue architecture to improve ablation outcomes by enabling targeted ablation based on tissue structure.
Frank Provenzano: Neuroimaging Informatics Informed Quantitative Measures
Neuroimaging has long been a staple in the diagnostic toolbox of neurologists across suspected diseases and specialties, with new ways to interpret existing clinically acquired imaging with both simple and complex methods. Conditions with no established neuroimaging signatures are being re-examined, including regard for heterogeneity and image quality, with new AI-driven imaging methods, revealing identifiable quantitative measures that may have extend the clinical utility of “already acquired” neuroimaging and inform pathogenic processes. Here we will explore some of these methods across disease and discipline, including Alzheimer’s Disease, cognitive aging, Lewy Body Dementia, epilepsy and others, and how one can capitalize on available open data neuroimaging to better understand disease risk and progression, potential therapeutics and surgical guidance.
