Abstract
This dissertation has introduced several methodological contributions that are required to enable nuclear image guidance in the interventional suite. The contributions have primarily been evaluated for use in radioembolization since this treatment is one of the most promising candidates to benefit from interventional nuclear imaging.
Interventional Scanner
The first task to enable interventional nuclear imaging is to develop a compact and flexible scanner with an open gantry that can also provide an anatomical reference for the nuclear images. A design was proposed and evaluated with a prototype scanner in CHAPTER 2 and with a full-scale device in CHAPTER 3. Metrics that are relevant for radioembolization were accurately retrieved from the images.
Instead of developing a new scanner, it might also be possible to adjust the scanner that is already available in the intervention room. In CHAPTER 4, it was demonstrated that the x-ray flat panel detector of the CBCT scanner can be used for nuclear imaging. The reconstruction quality of a conventional gamma camera was approached.
Motion Compensation
The nuclear reconstruction should be accurate if the distribution is used for the planning of radionuclide therapies. Patient respiratory motion is a major degrading factor for the reconstruction accuracy and should hence be compensated. Conventional methods to compensate for respiratory motion require external apparatus or high activity levels. Both are undesired in an interventional setting. A new method to compensate for respiratory motion was proposed by making use of the simultaneous acquisition of nuclear and fluoroscopic images in the interventional scanner. The method was evaluated with simulations in CHAPTER 5 and with a phantom study in CHAPTER 6.
Faster Scanning
The interventional scan duration should be as short as possible in order not to considerably prolong the time spent in the intervention room. Converging collimators (e.g. the cone beam collimator and the multifocal collimator) can focus on a specific organ to achieve higher sensitivity and hence could be used to shorten the scan duration. Quantitative reconstruction software for such collimators was developed and evaluated with digital simulations in CHAPTER 7. The quantitative reconstructor of the multifocal collimator was validated with phantom experiments in CHAPTER 8.
A different method of shortening the interventional scan duration, by evaluating the radionuclide distribution during scanning, was proposed in CHAPTER 9. Multiple fast scans of one minute each were made and the reconstruction was updated with the newly obtained counts after every rotation and the scan was hence terminated when the acquired images were sufficiently stable.
Faster Reconstruction
The nuclear reconstructions should be available quickly after scanning to ensure that the valuable time in the intervention room does not go to waste. Current state-of-the-art reconstructors require several minutes to complete. In CHAPTER 10, a different method for reconstruction was explored. This approach first reconstructed the projections using filtered back projection and then passed the results to an image enhancement convolutional neural network. The deep learning reconstruction method provided images with a quality similar to that of state-of-the-art reconstruction but completed within seconds.
Interventional Scanner
The first task to enable interventional nuclear imaging is to develop a compact and flexible scanner with an open gantry that can also provide an anatomical reference for the nuclear images. A design was proposed and evaluated with a prototype scanner in CHAPTER 2 and with a full-scale device in CHAPTER 3. Metrics that are relevant for radioembolization were accurately retrieved from the images.
Instead of developing a new scanner, it might also be possible to adjust the scanner that is already available in the intervention room. In CHAPTER 4, it was demonstrated that the x-ray flat panel detector of the CBCT scanner can be used for nuclear imaging. The reconstruction quality of a conventional gamma camera was approached.
Motion Compensation
The nuclear reconstruction should be accurate if the distribution is used for the planning of radionuclide therapies. Patient respiratory motion is a major degrading factor for the reconstruction accuracy and should hence be compensated. Conventional methods to compensate for respiratory motion require external apparatus or high activity levels. Both are undesired in an interventional setting. A new method to compensate for respiratory motion was proposed by making use of the simultaneous acquisition of nuclear and fluoroscopic images in the interventional scanner. The method was evaluated with simulations in CHAPTER 5 and with a phantom study in CHAPTER 6.
Faster Scanning
The interventional scan duration should be as short as possible in order not to considerably prolong the time spent in the intervention room. Converging collimators (e.g. the cone beam collimator and the multifocal collimator) can focus on a specific organ to achieve higher sensitivity and hence could be used to shorten the scan duration. Quantitative reconstruction software for such collimators was developed and evaluated with digital simulations in CHAPTER 7. The quantitative reconstructor of the multifocal collimator was validated with phantom experiments in CHAPTER 8.
A different method of shortening the interventional scan duration, by evaluating the radionuclide distribution during scanning, was proposed in CHAPTER 9. Multiple fast scans of one minute each were made and the reconstruction was updated with the newly obtained counts after every rotation and the scan was hence terminated when the acquired images were sufficiently stable.
Faster Reconstruction
The nuclear reconstructions should be available quickly after scanning to ensure that the valuable time in the intervention room does not go to waste. Current state-of-the-art reconstructors require several minutes to complete. In CHAPTER 10, a different method for reconstruction was explored. This approach first reconstructed the projections using filtered back projection and then passed the results to an image enhancement convolutional neural network. The deep learning reconstruction method provided images with a quality similar to that of state-of-the-art reconstruction but completed within seconds.
Original language | English |
---|---|
Awarding Institution |
|
Supervisors/Advisors |
|
Award date | 17 Sept 2020 |
Publisher | |
Print ISBNs | 978-90-393-7315-6 |
DOIs | |
Publication status | Published - 17 Sept 2020 |
Keywords
- Nuclear imaging
- Interventional radiology
- Hybrid imaging
- SPECT
- CBCT
- Radioembolization