Opportunistic Screening for Pancreatic Cancer using Computed Tomography Imaging and Radiology Reports

Presentation Time: 03:15 PM - 03:27 PM

Abstract Keywords: Artificial Intelligence, Bioinformatics, Deep Learning, Machine Learning, Natural Language Processing, Clinical Decision Support
Primary Track: Foundations
Programmatic Theme: Clinical Research Informatics

Pancreatic ductal adenocarcinoma (PDAC) is a highly aggressive cancer, with most cases diagnosed at stage IV and a five-year overall survival rate below 5%. Early detection and prognosis modeling are crucial for improving patient outcomes and guiding early intervention strategies. In this study, we developed and evaluated a deep learning fusion model that integrates radiology reports and CT imaging to predict PDAC risk. The model achieved a concordance index (C-index) of 0.6750 (95% CI: 0.6429, 0.7121) and 0.6435 (95% CI: 0.6055, 0.6789) on the internal and external dataset, respectively, for 5-year survival risk estimation. Kaplan-Meier analysis demonstrated significant separation (p<0.0001) between the low and high risk groups predicted by the fusion model. These findings highlight the potential of deep learning-based survival models in leveraging clinical and imaging data for pancreatic cancer.

Speaker:
David Le, PhD
Mayo Clinic

Authors:
Ramon Correa-Medero, B.S. - Mayo Clinic; Amara Tariq, Ph.D. - Mayo Clinic Arizona; Bhavik Patel, MD, MBA - Mayo Clinic; Motoyo Yano, MD, PhD - Mayo Clinic; Imon Banerjee, PhD - Arizona State U, Mayo Clinic;

---

OMT-SAM: A Multi-Modal, Text-Guided Framework for Accurate Abdominal Organ Segmentation in Medical Imaging

Presentation Time: 03:27 PM - 03:39 PM

Abstract Keywords: Imaging Informatics, Deep Learning, Large Language Models (LLMs)
Primary Track: Foundations

Accurate segmentation is vital for treatment planning and disease monitoring. Existing methods mainly rely on unimodal inputs, requiring manual annotations. Medical imaging often captures multiple organs in a single scan, complicating segmentation. To improve this, MedSAM, based on the Segment Anything Model (SAM), enhances segmentation using image features with user prompts. We propose OMT-SAM, incorporating CLIP encoders and multi-scale visual features to address limitations and improve segmentation accuracy. Evaluations show OMT-SAM outperforms MedSAM with a Dice Similarity Coefficient of 0.937.

Speaker:
Jiancheng Ye, PhD
Weill Cornell Medicine

Author:
Wenjie Zhang, Master - Weil Cornell Medicine;

---

Automated Image Registration Method for In Vivo Confocal Microscopy of the Corneal Sub-basal Nerve Plexus

Presentation Time: 03:39 PM - 03:51 PM

Abstract Keywords: Imaging Informatics, Quantitative Methods, Information Extraction
Primary Track: Applications
Programmatic Theme: Clinical Informatics

In vivo confocal microscopy (IVCM) assesses corneal innervation in the sub-basal nerve plexus but is typically quantified manually from a single z-scan, limiting biomarker extrapolation for diagnosing limbal stem cell deficiency (LSCD). We developed an automated 3D reconstruction method of IVCM image volumes to improve sub-basal nerve density quantification. Our dataset comprised 99 IVCM stacks from 63 LSCD eyes (51 patients) and 23 stacks from 15 normal eyes. We designed an image registration algorithm combining phase correlation and homography transformation, which achieved a pairwise image correlation of 0.69 and mutual information of 0.60, significantly outperforming manual registration (0.60 and 0.43, respectively; p<0.001). Validation on an independent dataset of 325 volume scans from 24 eyes of 12 unilateral, severe LSCD patients yielded a correlation of 0.75 and MI of 0.76. This method enhances sequential IVCM scan alignment and supports more accurate, reproducible 3D evaluation of LSC biomarkers.

Speaker:
Nathan Siu, B.S.
University of California, Los Angeles

Authors:
Nathan Siu, B.S. - University of California, Los Angeles; Micah Vinet, MS - UCLA; Parth Shettiwar, MS - University of California, Los Angeles; Thai Tran, B.S. - University of California, Los Angeles; Kyle Chen, BS - University of California, Los Angeles; Theo Stoddard-Bennett, MD - University of California, Los Angeles; Clemence Bonnet, MD, PhD - University of California, Los Angeles; Corey Arnold, PhD - UCLA; William Speier, PhD - UCLA; Sophie Deng, MD, PhD - University of California, Los Angeles;

---

CT-Bench: A Comprehensive Benchmark Dataset for Multimodal AI in CT Analysis

Presentation Time: 03:51 PM - 04:03 PM

Abstract Keywords: Artificial Intelligence, Large Language Models (LLMs), Diagnostic Systems, Machine Learning
Primary Track: Foundations
Programmatic Theme: Translational Bioinformatics

AI can detect lesions in CT scans and pre-fill radiology reports, but progress is limited by the lack of publicly available, well-annotated CT datasets. To address this, we introduce CT-Bench, a benchmark dataset with two components: Lesion Image & Metadata Set (20,335 lesion images with descriptions, sizes, and bounding boxes) and QA Benchmark (2,850 question-answer pairs). Fine-tuning on CT-Bench significantly improved AI performance, boosting BiomedCLIP's accuracy from 41.6% to 61.8%. This dataset enhances lesion analysis, improving efficiency in radiology workflows.

Speaker:
Qingqing Zhu, Ph.D
National Institute of Health

Authors:
Qingqing Zhu, PHD - National Institutes of Health; Qiao Jin, M.D. - National Institutes of Health; Tejas Sudharshan Mathai, Ph.D - National Institute of Health; Yin Fang, Ph.D. - National Institute of Health; Zhizheng Wang, Ph.D - National Institutes of Health; Yifan Yang, B.S. - NCBI, NLM/NIH; Maame Gyamfi, MD - Howard University College of Medicine; Benjamin Hou, Ph.D. - National Institute of Health; Ran Gu, Ph.D. - National Institute of Health; Praveen Thoppey Srinivasan Balamuralikrishna, M.D - National Institute of Health; Ronald M. Summers, M.D., Ph.D. - National Institute of Health; Zhiyong Lu, PhD - National Library of Medicine, NIH;

---

Fair Multi-modal Canonical Correlation Analysis: A Neuroimaging Study of Alzheimer's Disease

Presentation Time: 04:03 PM - 04:15 PM

Abstract Keywords: Fairness and Elimination of Bias, Machine Learning, Bioinformatics
Primary Track: Foundations

This study addresses fairness concerns in Multi-modal Canonical Correlation Analysis (MCCA), a technique for analyzing relationships across multiple datasets. We introduce Fair MCCA (F-MCCA), which mitigates bias by optimizing for both correlation performance and demographic fairness. Our method quantifies disparities using Correlation Disparity Error (CDE) and employs a multi-objective optimization framework to derive projection matrices that achieve consistent correlation levels across sensitive groups. We validate F-MCCA on neuroimaging data from the Alzheimer's Disease Neuroimaging Initiative using sex as the sensitive attribute. Experiments demonstrate that F-MCCA substantially improves fairness metrics with minimal correlation performance sacrifice. In downstream classification tasks, F-MCCA reduces demographic parity difference and equalized odds difference while maintaining comparable accuracy to standard approaches. Results confirm that our method effectively balances analytical performance with fairness considerations, supporting more unbiased healthcare applications of multi-modal data analysis.

Speaker:
Zhuoping Zhou, Master of Art
University of Pennsylvania

Authors:
Zhuoping Zhou, Master of Art - University of Pennsylvania; Boning Tong, MSE - University of Pennsylvania; Bojian Hou, PhD - University of Pennsylvania; Christos Davatzikos, PhD - University of Pennsylvania; Qi Long, Ph.D. - University of Pennsylvania; Li Shen, Ph.D. - University of Pennsylvania;

---

A global study of alternative splicing in non-small cell lung cancer identifies unique histological and population-specific differences

Presentation Time: 04:15 PM - 04:27 PM

Abstract Keywords: Bioinformatics, Computational Biology, Data Mining, Health Equity, Omics (genomics, metabolomics, proteomics, transcriptomics, etc.) and Integrative Analyses, Personal Health Informatics, Precision Medicine
Working Group: Genomics and Translational Bioinformatics Working Group
Primary Track: Applications
Programmatic Theme: Translational Bioinformatics

Lung cancer is more prevalent in African American (AA) men with higher incidence and mortality rates compared to European-American (EA) men. This study identified novel alternative splicing (AS) events in lung tumors from patients with adenocarcinoma (LUAD) and squamous cell carcinoma (LUSC). The research uncovered tumor subtype- and population-specific AS events linked to oncogenic pathways and cancer driver genes, such as SYNE2, CD44, and AFDN. These findings offer insights into molecular differences between AA and EA populations, potentially advancing therapeutic approaches and health equity.

Speaker:
Saman Zeeshan, PhD
The University of Missouri, Kansas City

Authors:
Saman Zeeshan, PhD - The University of Missouri, Kansas City; Bhavik Dalal, PhD - National Cancer Institute; Rony Arauz, PhD - National Cancer Institute; Adriana Zingone, PhD - National Cancer Institute; Curtis Harris, PhD - National Cancer Institute; Hossein Khiabanian, PhD - Rutgers Cancer Institute of New Jersey; Sharon Pine, PhD - University of Colorado Anschutz Medical Campus; Brid Ryan, PhD - National Cancer Institute;

---