The development of an in silico imaging pipeline for the simulation and deep learning-based tumour segmentation of spectral photon-counting CT brain images
Loading...
Embargo End Date
2026-02-20
ICR Authors
Authors
Kattau, M
Document Type
Thesis or Dissertation
Date
2025-02-20
Date Accepted
Abstract
Photon-counting CT (PCCT) is an emerging technology employing photon-counting detectors instead of the energy-integrating detectors (EIDs) used in conventional CT. Previous studies have shown benefits of PCCT for neuroimaging, including improved contrast to noise ratio (CNR) and soft tissue contrast. The impact of these advancements on radiotherapy treatment planning is currently unclear and the unmet need this work aims to address. The increased soft tissue contrast makes PCCT a promising candidate for a single-modality approach in stereotactic radiosurgery (SRS). Current clinical practice relies on an MRI scan for soft tissue differentiation in addition to the conventional CT required for dose calculation, however PCCT could achieve both aims with a single scan. This work assesses the feasibility of using PCCT for tumour delineation in SRS.
Recent studies have also shown the benefit of deep learning-based tumour segmentation for treatment planning as opposed to manual delineation. Therefore, the potential of using PCCT as single modality in combination with deep learning for SRS treatment planning is evaluated. An in-silico imaging pipeline is developed simulating PCCT brain images based on computational phantoms derived from clinical data. The images are used as input to a deep learning model for tumour segmentation.
The results demonstrate improved CNR within the tumour region on the PCCT images compared to their EID counterparts, showing the potential of a single-modality approach. Moreover, PCCT image quality was more robust towards an increase in electronic noise. Automatic tumour segmentation on the simulated images is feasible, but a larger number of patients is required to determine the impact of the improved image quality on segmentation performance.
The developed in-silico pipeline is of great benefit to the research community, as it offers a cost- and time-effective way to evaluate feasibility and methodologies of research studies on the implementation of PCCT in neuro-oncology and beyond.
Citation
2025
DOI
Source Title
Publisher
Institute of Cancer Research (University Of London)
ISSN
eISSN
Collections
Research Team
Multimodal Mol Imaging