Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Leveraging Multiphase CT for Quality Enhancement of Portal Venous CT: Utility for Pancreas Segmentation

Xinya Wang, Tejas Sudharshan Mathai, Boah Kim, Ronald M. Summers

TL;DR

This study introduces a 3D progressive fusion and non-local (PFNL) network to leverage three CT phases—non-contrast, arterial, and portal venous—to enhance the portal venous phase, addressing variability from low-dose protocols and artifacts. The approach is trained with synthetic degradations to mimic low-quality inputs and optimized with a combined $L_1$ and $L_1$ Sobel-edge loss, enabling sharper structural boundaries. A pancreas segmentation proxy using TotalSegmentator demonstrates that multiphase-enhanced PV CT yields meaningful improvements in Dice and NSD over degraded scans, with 3D-PFNL outperforming a single-phase baseline in boundary preservation and segmentation stability. Despite a small test cohort, results suggest practical value in improving downstream tasks and generalizing to other phases, motivating larger-scale validation and extension to additional organs. The work is supported by NIH resources and relies on publicly available datasets for ethical compliance.

Abstract

Multiphase CT studies are routinely obtained in clinical practice for diagnosis and management of various diseases, such as cancer. However, the CT studies can be acquired with low radiation doses, different scanners, and are frequently affected by motion and metal artifacts. Prior approaches have targeted the quality improvement of one specific CT phase (e.g., non-contrast CT). In this work, we hypothesized that leveraging multiple CT phases for the quality enhancement of one phase may prove advantageous for downstream tasks, such as segmentation. A 3D progressive fusion and non-local (PFNL) network was developed. It was trained with three degraded (low-quality) phases (non-contrast, arterial, and portal venous) to enhance the quality of the portal venous phase. Then, the effect of scan quality enhancement was evaluated using a proxy task of pancreas segmentation, which is useful for tracking pancreatic cancer. The proposed approach improved the pancreas segmentation by 3% over the corresponding low-quality CT scan. To the best of our knowledge, we are the first to harness multiphase CT for scan quality enhancement and improved pancreas segmentation.

Leveraging Multiphase CT for Quality Enhancement of Portal Venous CT: Utility for Pancreas Segmentation

TL;DR

This study introduces a 3D progressive fusion and non-local (PFNL) network to leverage three CT phases—non-contrast, arterial, and portal venous—to enhance the portal venous phase, addressing variability from low-dose protocols and artifacts. The approach is trained with synthetic degradations to mimic low-quality inputs and optimized with a combined and Sobel-edge loss, enabling sharper structural boundaries. A pancreas segmentation proxy using TotalSegmentator demonstrates that multiphase-enhanced PV CT yields meaningful improvements in Dice and NSD over degraded scans, with 3D-PFNL outperforming a single-phase baseline in boundary preservation and segmentation stability. Despite a small test cohort, results suggest practical value in improving downstream tasks and generalizing to other phases, motivating larger-scale validation and extension to additional organs. The work is supported by NIH resources and relies on publicly available datasets for ethical compliance.

Abstract

Multiphase CT studies are routinely obtained in clinical practice for diagnosis and management of various diseases, such as cancer. However, the CT studies can be acquired with low radiation doses, different scanners, and are frequently affected by motion and metal artifacts. Prior approaches have targeted the quality improvement of one specific CT phase (e.g., non-contrast CT). In this work, we hypothesized that leveraging multiple CT phases for the quality enhancement of one phase may prove advantageous for downstream tasks, such as segmentation. A 3D progressive fusion and non-local (PFNL) network was developed. It was trained with three degraded (low-quality) phases (non-contrast, arterial, and portal venous) to enhance the quality of the portal venous phase. Then, the effect of scan quality enhancement was evaluated using a proxy task of pancreas segmentation, which is useful for tracking pancreatic cancer. The proposed approach improved the pancreas segmentation by 3% over the corresponding low-quality CT scan. To the best of our knowledge, we are the first to harness multiphase CT for scan quality enhancement and improved pancreas segmentation.
Paper Structure (15 sections, 1 equation, 5 figures, 2 tables)

This paper contains 15 sections, 1 equation, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Patient Sample
  4. Reference Standard
  5. Data Preprocessing
  6. Model for Portal Venous Phase Quality Enhancement
  7. Pancreas Segmentation
  8. Experiments
  9. Statistical Analysis
  10. Results
  11. Quality Enhancement Results
  12. Segmentation Results
  13. Discussion
  14. ACKNOWLEDGMENTS
  15. Compliance with Ethical Standards

Figures (5)

  • Figure 1: The overall framework for portal venous scan quality improvement is shown. Three degraded and low-quality CT phases (non-contrast, arterial, and portal venous) were used to train a 3D-PFNL quality enhancement network. At test time, the model generated a high-quality portal venous CT volume, given the three inputs. The effect of scan quality enhancement was assessed with the public TotalSegmentator tool that segmented the pancreas in a proxy segmentation task.
  • Figure 2: Qualitative examples of low-quality (LQ) portal venous (PV) CT, 3D-RCAN restored PV CT and 3D-PFNL restored PV CT, respectively. Orange arrows indicate the sharper and clearer edges of the pancreas by the proposed 3D-PFNL quality enhancement approach.
  • Figure 3: Dice score
  • Figure 4: Normalized Surface Distance
  • Figure 6: Segmentation results of TotalSegmentator (TS) on the reference (original) portal venous (PV) CT scan, low-quality (LQ) PV CT, 3D-RCAN restored PV CT and 3D-PFNL restored PV CT, respectively. Manual pancreas annotations are in yellow, while the automated segmentations by TS are in red.