Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep PCCT: Photon Counting Computed Tomography Deep Learning Applications Review

Ana Carolina Alves, André Ferreira, Gijs Luijten, Jens Kleesiek, Behrus Puladi, Jan Egger, Victor Alves

TL;DR

PCCT addresses core CT limitations by counting individual photons and exploiting energy-resolved data, enabling improved contrast-to-noise and potential dose reductions. The paper surveys how deep learning and radiomics are applied to PCCT, comparing PCCT to EICT and detailing workflows, reconstruction architectures, and results across reconstruction, segmentation, artifact mitigation, and material decomposition. Evidence indicates PCCT enhances image quality and iodine detectability, and radiomic features can be more discriminative, though some features shift with energy-level reconstructions; deep learning further improves reconstruction, denoising, and decomposition, but generalization and standardization remain key obstacles. Overall, the integration of DL with PCCT holds promise for high-resolution, low-dose, spectral imaging and the concept of a virtual biopsy, contingent on broader data sharing, standardized metrics, and regulatory alignment.

Abstract

Medical imaging faces challenges such as limited spatial resolution, interference from electronic noise and poor contrast-to-noise ratios. Photon Counting Computed Tomography (PCCT) has emerged as a solution, addressing these issues with its innovative technology. This review delves into the recent developments and applications of PCCT in pre-clinical research, emphasizing its potential to overcome traditional imaging limitations. For example PCCT has demonstrated remarkable efficacy in improving the detection of subtle abnormalities in breast, providing a level of detail previously unattainable. Examining the current literature on PCCT, it presents a comprehensive analysis of the technology, highlighting the main features of scanners and their varied applications. In addition, it explores the integration of deep learning into PCCT, along with the study of radiomic features, presenting successful applications in data processing. While acknowledging these advances, it also discusses the existing challenges in this field, paving the way for future research and improvements in medical imaging technologies. Despite the limited number of articles on this subject, due to the recent integration of PCCT at a clinical level, its potential benefits extend to various diagnostic applications.

Deep PCCT: Photon Counting Computed Tomography Deep Learning Applications Review

TL;DR

PCCT addresses core CT limitations by counting individual photons and exploiting energy-resolved data, enabling improved contrast-to-noise and potential dose reductions. The paper surveys how deep learning and radiomics are applied to PCCT, comparing PCCT to EICT and detailing workflows, reconstruction architectures, and results across reconstruction, segmentation, artifact mitigation, and material decomposition. Evidence indicates PCCT enhances image quality and iodine detectability, and radiomic features can be more discriminative, though some features shift with energy-level reconstructions; deep learning further improves reconstruction, denoising, and decomposition, but generalization and standardization remain key obstacles. Overall, the integration of DL with PCCT holds promise for high-resolution, low-dose, spectral imaging and the concept of a virtual biopsy, contingent on broader data sharing, standardized metrics, and regulatory alignment.

Abstract

Medical imaging faces challenges such as limited spatial resolution, interference from electronic noise and poor contrast-to-noise ratios. Photon Counting Computed Tomography (PCCT) has emerged as a solution, addressing these issues with its innovative technology. This review delves into the recent developments and applications of PCCT in pre-clinical research, emphasizing its potential to overcome traditional imaging limitations. For example PCCT has demonstrated remarkable efficacy in improving the detection of subtle abnormalities in breast, providing a level of detail previously unattainable. Examining the current literature on PCCT, it presents a comprehensive analysis of the technology, highlighting the main features of scanners and their varied applications. In addition, it explores the integration of deep learning into PCCT, along with the study of radiomic features, presenting successful applications in data processing. While acknowledging these advances, it also discusses the existing challenges in this field, paving the way for future research and improvements in medical imaging technologies. Despite the limited number of articles on this subject, due to the recent integration of PCCT at a clinical level, its potential benefits extend to various diagnostic applications.
Paper Structure (16 sections, 3 figures)

This paper contains 16 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Research Questions
  3. Search Strategy
  4. Manuscript outline
  5. EICT and PCCT Systems
  6. Radiomics Studies on PCCT images
  7. Radiomics Workflow
  8. Difference in Radiomic Analysis between PCCT and EICT
  9. Radiomics Limitations
  10. Deep learning applications in PCCT
  11. Image Reconstruction Techniques
  12. Segmentation Applications
  13. Reduction of artifacts and noise
  14. Material Decomposition
  15. The future of the PCCT
  16. ...and 1 more sections

Figures (3)

  • Figure 1: Selection of the papers according to PRISMA diagram Prisma.
  • Figure 2: Schematic drawing of Photon Counting Detector. Adapted from Meloni2023_review.
  • Figure 3: Representative axial CT slices at the acromioclavicular joint level showcase image quality variations across six combinations of dose protocols and detector technologies. The upper row displays full-dose scan protocols (10.0 mGy) for EICT, non-UHR-PCCT, and UHR-PCCT, while the lower row exhibits low-dose scan protocols (5.0 mGy) for the same technologies. Notably, lower doses in EICT reveal increased image noise, potentially impacting the assessability of cancellous bone structures. Retrieved from Patzer2023_nature.