Excellent lung cancer imaging and tumour identification using next generation computed tomography

Project Information

EXCITING-CT

Grant agreement ID: 750564

DOI

10.3030/750564

Project closed

EC signature date 27 February 2017

Start date 1 September 2017

End date 31 August 2019

Funded under

• EXCELLENT SCIENCE - Marie Skłodowska-Curie Actions

Total cost

€ 173 857,20

EU contribution

€ 173 857,20

Coordinated by

KUNGLIGA TEKNISKA HOEGSKOLAN
Sweden

Periodic Reporting for period 1 - EXCITING-CT (Excellent lung cancer imaging and tumour identification using next generation computed tomography)

Reporting period: 2017-09-01 to 2019-08-31

Summary of the context and overall objectives of the project

Lung cancer is the most common and deadly type of cancer globally, and survival is strongly linked to early detection. Therefore, it is highly desirable to screen high-risk individuals, and screening using computed tomography (CT) is currently the only method shown to improve survival. However, such screening is generally not carried out due to a high rate of false positives and increased patient x-ray exposure.

In this project we wanted to investigate the feasibility of using a CT system equipped with a unique, photon-counting silicon detector in screening for lung cancer. Advantages of the new detector include higher the resolution, no hard lower limit on the x-ray dose, reduced image artefacts, and higher contrast sensitivity compared to current CT detectors. Our hypothesis was that the patient x-ray dose can be significantly reduced in both lung cancer detection and characterisation. In addition, we believe that the higher resolution could reduce false positives, which would further reduce the dose through fewer follow-ups as well as lower the clinical costs and workload in screening. The use of photon-counting CT thus has the potential to simultaneously address both the main limitations currently making lung cancer screening infeasible.

Due to delays and unforeseen circumstances, scanning patients with lung cancer was not possible during the project. Instead, the potential clinical benefit of the improved resolution offered by the system was evaluated for head and neck scans. The main conclusions from the project were that the developed prototype system was suitable for patient scanning, offered better resolution than current hospital scanners, and that the improved resolution was deemed by clinicians to have clinical value in head and neck scans.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

- Subject-specific introductions and reading up on previous research
- Design of geometric calibration procedure for photon-counting detector when mounted in gantry, including design of necessary hardware and software
- Design of phantom for basis material decomposition calibration
- Hands-on training running a clinical CT gantry, including CT safety course
- Initial training on using the existing framework for running Monte Carlo simulations in preparation for future simulations of images from the CT system
- Networking with clinicians at the collaborating university hospital to ensure correctness and clinically relevancy of planned patient studies
- Four-day knowledge exchange and informal training with engineer from major CT manufacturer
- Three-day hands-on course on iterative reconstruction with the ODL-framework using preliminary phantom data from prototype CT scanner
- Supervision bachelor thesis projects on photon-counting CT. Two projects, five students in total
- Initial imaging of line-pair phantom showing improved spatial resolution compared to current state-of-the-art
- Initial imaging of anatomical skull phantom and corresponding (non-spectral) reconstruction of images to demonstrate clinical value of higher resolution
- Oral presentation “First evaluation of new photon-counting CT technology” at European Congress of Radiology 2018 in Vienna. Presentation resulted in “Best Scientific Paper Presentation Award 2018” within Physics of Medical Imaging.
- News article about research project and photon-counting CT published on the University’s external web page
- Development of semi-automated workflow for material decomposition calibration
- Improvement and simplification of geometric calibration routine
- Investigation into neural network solutions for material decomposition in projection space
- Outreach aimed at pupils in years 9 to 11 to talk about medical imaging, the role of the scientist and the current research project as part of the local European Researchers’ Night event
- Seminar talk describing current research to other scientists working with X-ray physics at home university
- Outreach aimed at non-academic staff from universities across Sweden participating in the 2019 National Exam Conference, talking about the current project and photon-counting CT in general
- Four-day workshop on deep learning in inverse problems, in particular the use of machine learning for reconstruction of CT images
- Installation of new small-focal-spot X-ray tube on prototype CT scanner with ensuing calibration
- Experiments and manuscript preparation for scientific article demonstrating the improved spatial resolution offered by the photon-counting CT prototype used in this project
- Initial phantom study together with clinical collaborators of the potential benefit of higher spatial resolution in stenosis imaging compared to a current state-of-the-art scanner
- Scan of sheep’s lungs and head together with clinical collaborators for evaluation of potential resolution benefit of prototype photon-counting scanner in clinical lung and head cases
- Half-semester university course in deep learning methods with the aim to use the acquired knowledge for pileup correction in order to improve the quality of the images produced by the prototype CT scanner
- Investigation and experiments into using machine learning methods for spectral pileup correction in order to improve the quality of the images from the prototype CT scanner

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

Between September and November 2017, there has been no progress beyond state of the art.

From December 2017 until July 2018, the main aim was to prepare the prototype CT system for the first imaging scans and to develop the necessary framework for image reconstruction and routines for patient imaging. Progress beyond current state of the art was made in two ways. First, it was shown that the photon-counting silicon detector developed by the research group enables higher spatial resolution than CT scanners that are common in hospitals today. The results were presented at the European Congress of Radiology and awarded best oral presentation within Physics in Medical Imaging. Second, a first ever CT scan of the head and neck of a real patient was carried out using the prototype CT system. This was likely the first scan of a human patient using a full-size CT scanner equipped with a silicon-based, photon-counting detector.

Between August 2018 and March 2019, the focus was to evaluate the improved resolution offered by the prototype scanner with a new small-focal-spot X-ray tube installed. However, after a second patient scan, it was decided to pause patient scanning before moving on to other clinical indications. This forced a deviation from the initial project plan and shifted the focus to characterising the benefits of higher spatial resolution. Potential for improvement beyond the state of the art was seen when comparing images of a stenosis phantom with those taken using a clinical scanner. In particular, the higher image resolution allowed for better distinction between different grades of stenosis in cases where little or no contrast agent was present, which could improve patient diagnostics.

From April to August 2019, work was aimed at investigating the use of machine learning to improve the spectral imaging of the system, which would be necessary for example in better lung nodule characterisation. Progress beyond state of the art was achieved by showing that the machine learning-based method performed better than the other approaches evaluated in parallel.

report-exciting-ct.jpg