Icon Legend

This session is not in your schedule.

This session is in your schedule. Click again to remove it.

Biomedical Imaging and Instrumentation

Biomedical Imaging and Instrumentation - Poster Session E

Z32 - Quantitative Perfusion Imaging from Non-Contrast 4D micro-CT for Pulmonary Embolism Evaluation in Preclinical Models

Saturday, October 11, 2025

10:00 AM - 11:00 AM PDT

Location: Exhibit Hall F, G & H

Presenting Author(s)

DC

Debarghya Chaki

Research Assistant
The University of Texas at Austin, United States

Co-Author(s)

Hsu-Ting Kuo, BS

PhD Candidate
University of Texas at Austin
Austin, Texas, United States
AB

Allan Barcena

Postdoctoral Fellow
MD Anderson Cancer Center, United States
HN

Huan Nguyen

Undergraduate Student
University of Texas at Austin, United States
DB

Dominic Bolinas

Graduate Student
MD Anderson Cancer Center, Texas, United States
MB

Marvin Bernardo

Research Data Coordinator
MD Anderson Cancer Center, United States

Last Author(s)

MM

Marites Melancon

Professor
MD Anderson Cancer Center, United States

Primary Investigator(s)

Edward Castillo, PhD

Associate Professor & Associate Chair
The University of Texas at Austin
Austin, Texas, United States

Introduction: : Up to 900,000 people may be annually affected by pulmonary embolism (PE), a condition in which a blood clot occludes the pulmonary artery often leading to pain or difficulty breathing, and in 25% of cases, sudden death as the first symptom according to the Centers for Disease Control and Prevention. The current standard for PE diagnosis and imaging is chest computed tomography angiography (CTA) largely for its accessibility and simplicity. CTA is however often inaccessible to patients during emergencies due to allergic reaction against the iodine based contrast agent or renal insufficiency. For such patients, there is a need for alternative diagnosis methods for PE. While alternate methods including single photon emission computed tomography (SPECT), contrast-enhanced magnetic resonance angiography, and non-contrast MRA are available, they each include drawbacks regarding cost, time, or specialized equipment making them unsuitable in emergency settings.

With CT-derived perfusion, pulmonary perfusion can be quantified without the need for contrast agents. This method requires mathematical modeling and computing on inhale/exhale image pairs from four-dimensional CT (4DCT). Model outputs are used to identify regions associated with PE based on perfusion scores across regions. In this study, we verify the performance and localization potential of our method using a mouse model to assess PE diagnostic and evaluation criteria across the lung. The goal of this study is to further the development of a method that may be employed in emergency department settings for safe and accurate PE diagnosis.

Materials and

Methods: : Our model includes a scoring scheme based on optimally derived CT-perfusion (CT-P) thresholds extracted from human 4DCT. These thresholds were determined using Particle Swarm Optimization, refined via a formula to best differentiate between positive and negative cases once CT-P values were obtained for each case. Data is preprocessed by converting CT Hounsfield units into density values, and pulmonary blood mass change is calculated through an automatic segmentation pipeline per case. This mass change is calculated as the absolute value of the difference of the inhale and exhale masses, and classification is determined through optimizing normal threshold vs. perfusion deficit. For this study, the method was adjusted to use manual segmentation and simplified perfusion mapping for the mouse model.

We conducted an IACUC-approved imaging study with three mice that underwent pulmonary embolization followed by non-contrast 4D micro-CT imaging. Inhale/exhale CT image pairs were acquired both before and after embolization to assess regional changes in lung perfusion resulting from the induced occlusions. To quantify these changes, we employed a rigid registration pipeline that aligned inhalation-phase images to their corresponding exhalation-phase volumes, enabling voxelwise comparison of pulmonary tissue density shifts. This registration facilitated the calculation of perfusion scores across the respiratory cycle and allowed for precise spatial visualization of perfusion deficits in affected lung regions. Statistical analysis was then performed with a Wilcoxon signed-rank test.

Results, Conclusions, and Discussions:: In three murine embolism models, we applied our non‐contrast 4DCT‑derived CT‑Perfusion (CT‑P) pipeline, involving inhale‑to‑exhale rigid registration followed by voxelwise perfusion scoring to quantify lung‐specific blood flow before and after pulmonary embolization. Quantitatively, we measured CT‑P scores in the left (LL) and right (RL) lungs of each animal (n = 3), yielding six paired observations. A Wilcoxon signed‑rank test on the pooled lobe scores demonstrated a statistically significant reduction in perfusion post‑embolism (median pre = 0.0035, post = 0.0014; p = 0.031). Individually, all lungs showed consistent declines, though the small cohort precluded significance at the single‐lung level (Wilcoxon p = 0.250 for both LL and RL).

Qualitative visualization of perfusion maps confirmed these findings. Overlaid CT‑P colormaps clearly delineated regional perfusion deficits aligned with known embolus locations. In particular, Mouse 02 exhibited near‐complete washout in areas of its right lung (post/pre ratio ≈ 0.05), while Mouse 03 retained moderate perfusion (ratio ≈ 0.78), illustrating the method’s dynamic range and sensitivity to variable embolic burdens. Perfusion differences are observed in the figure below.

These results underscore several key points. First, the significant pooled test supports CT‑P as a robust, distribution‑free biomarker for detecting global perfusion losses due to embolism. Second, the per‑lung analysis, though underpowered, reveals biologically plausible heterogeneity in embolus impact, suggesting that larger cohorts would yield statistically significant lung‑specific changes. Third, the registration‑based color overlays offer an intuitive spatial readout allowing for rapid visual confirmation of findings during urgent situations.

Our non‑contrast 4DCT CT‑Perfusion workflow delivers quantitatively and visually compelling evidence of perfusion impairment following pulmonary embolism. The significant pooled perfusion drop (p < 0.05) validates CT‑P’s potential as a non‑invasive diagnostic metric, while the high‐resolution maps provide detailed localization of flow deficits. Future work will expand to larger animal cohorts and clinical 4DCT studies, aiming to establish CT‑P as a routine tool for early embolism detection and treatment monitoring. The animal model presented here provides a basis for larger scale validation aligned with the overall goal of our PE study, including incorporating the model with breath hold CT, improving model sensitivity, and clinical implementation in emergency departments.

Acknowledgements and/or References (Optional): :