From Pixels to Prognosis: AI, Imaging, and Alternative Physiological Signals for Chronic Obstructive Pulmonary Disease

#philadelphia, #october #ieee #AI #artificial-intelligence #medical-imaging #deep-learning
Share

IEEE Section Night Webinar

Webinar link and password will be forwarded to all registered participants after Noon on the day of the meeting. Check your spam folder if you don't see the email. 

Webinar is open to all.

From Pixels to Prognosis:
AI, Imaging, and Alternative Physiological Signals for Chronic Obstructive Pulmonary Disease

Part I  ·  AI and Quantitative Imaging for COPD Chronic obstructive pulmonary disease (COPD) is the third leading cause of death worldwide and remains underdiagnosed in more than half of affected individuals. Spirometry, the current diagnostic standard, is insensitive to early structural change, and quantitative information present in every chest CT scan is largely unexploited in routine radiology practice. This talk reviews a decade of work developing artificial intelligence and quantitative imaging methods that recover disease-relevant structural, mechanical, and textural information invisible to the human eye, drawing on validation in the COPDGene, SPIROMICS, and National Lung Screening Trial cohorts.

The talk covers three complementary computational threads: (i) geometry-derived CT biomarkers including airway fractal dimension, tapering, surface-area-to-volume ratio, and CT-derived dysanapsis, all independently associated with respiratory morbidity and mortality; (ii) silent zones of lung disease, where deep learning applied to visually normal parenchyma reveals subclinical radiomic and mechanical signatures that predict emphysema onset, exacerbation, and mortality years before conventional criteria trigger; and (iii) deep neural network analysis of raw spirometry flow-volume curves, which recovers CT-defined structural phenotype from a single breath and separates true ventilatory restriction from obstruction. The talk closes with translation of these methods through UAB spin-offs and where the field is moving with vision-language foundation models for multimodal chest imaging at the point of care.

Part II  ·  Deep Learning on the ECG for COPD Detection  This talk reviews recent evidence that convolutional neural networks can identify COPD directly from the 12-lead electrocardiogram (ECG), drawing on the study by Vaid et al. Because COPD produces measurable cardiopulmonary effects, the ECG, a ubiquitous, low-cost, and effort-independent signal, may carry latent information about airflow obstruction. The Mount Sinai CNN model was trained on 208,231 ECGs from 18,225 COPD cases matched to 49,356 controls and evaluated on an internal holdout, an external hospital cohort, and the UK Biobank, achieving AUC of 0.80, 0.82, and 0.75 respectively. The talk covers the model architecture, its linkage to concurrent spirometry, the explainability analyses identifying P-wave changes as the most informative ECG feature for COPD, and how ECG-based prediction could complement the imaging pipelines described in Part I and extend COPD screening into resource-limited settings.


  Date and Time

  Location

  Hosts

  Registration


  • Add_To_Calendar_icon Add Event to Calendar

Loading virtual attendance info...

  • Contact Event Host
  • Starts 25 September 2026 04:00 AM UTC
  • Ends 20 October 2026 04:00 PM UTC
  • Admission fee ?

  Speakers

Kelly Powell, MBA of University of Alabama at Birmingham

Topic:

From Pixels to Prognosis: AI, Imaging, and Alternative Physiological Signals for Chronic Obstructive Pulmonary Disease

Biography:

Kelly Powell, MBA, is Director of AI Initiatives at MHIBI at UAB Heersink School of Medicine, where she oversees graduate programs in AI in Medicine and Healthcare Innovation and supports AI adoption across the health system. She works closely with cross-functional teams across clinical, technical, and administrative departments to align AI efforts with institutional strategy and supports grant development. A Navy veteran and former Hospital Corpsman, she holds a BS in Psychology and an MBA, bringing a unique combination of clinical, strategic, and operational experience to healthcare AI.

Address:United States

Sandeep Bodduluri, MS, PhD of University of Alabama at Birmingham

Topic:

From Pixels to Prognosis: AI, Imaging, and Alternative Physiological Signals for Chronic Obstructive Pulmonary Disease

Biography:

Sandeep Bodduluri, PhD, is Assistant Professor of Medicine (Pulmonary, Allergy, and Critical Care) and of Electrical and Computer Engineering at UAB. He directs AI research at the Center for Lung Analytics and Imaging Research (CLAIR) and serves as founding Director of AI at the Marnix E. Heersink Institute for Biomedical Innovation (MHIBI), where he was appointed Deputy Director in 2026. He is Program Director of the MS in AI in Medicine and MS in Healthcare Innovation and Co-Director of the PhD in Biomedical and Health Informatics (AI track). His group develops imaging and AI methods for COPD, alpha-1 antitrypsin deficiency, non-tuberculous mycobacterial infection, TBI, and difficult airway; he has 71 peer-reviewed publications, three U.S. patents, and five UAB spin-offs, and serves as Associate Editor of Intelligence Based Medicine and the British Journal of Radiology COPD Special Issue.





Agenda

WEBINAR: 7:00 - 8:00 P.M. 

The Zoom Webinar link and password will be forwarded to all registered participants after Noon on the day of the meeting. Check your spam folder if you don't see the email. 

Webinar is open to all.

PDH certificates are available and an evaluation form will be emailed to you after the meeting. PDH certificate are sent by IEEE USA 3-4 weeks after the meeting.