Abstract
Background
CT can accurately measure left ventricular ejection fraction (LVEF) but with a high radiation dose. The potential for dose reduction in CT LVEF was explored and results used in designing a protocol to be validated clinically. A one-stop LVEF and chest-abdomen-pelvis (CAP) CT protocol was tested, for use in, e.g., in monitoring for chemotherapy-induced cardiotoxicity to potentially replace nuclear medicine MUGA scans.
Methods
Exploratory studies were performed to determine a lower radiation dose limit using 3D-printed heart phantoms and simulated dose reduction in clinical images. The phantoms were produced using 3D-printing of a clinical cardiac CT in both sys- and diastolic phases, and were scanned at multiple parameter combinations in an anthropormorphic chest phantom.
A series of clinical, functional cardiac CT scans were subjected to simulated noise reduction at 25, 10, 5 and 2 \% of the originally applied dose, using a custom-made algorithm. Following validation, LV volumetric measurements were performed at all dose levels and compared to full-dose scans with Bland-Altman plots.
From these results, a clinical protocol was designed.
The protocol was tested on 82 patients, with cardiac MRI as a reference. The protocol was combined with CAP CT with an adapted injection protocol in a subgroup of patients. 49 patients also had MUGA scans performed. Absolute agreement and classification accuracy for reduced LVEF<50\% were examined with Bland-Altman plots and ROC. CAP image quality was compared with previous scans using Visual Grading Characteristics. Inter-reader agreement was measured with ICC.
Finally, simulation of reduced temporal resolution and discontinuity artefacts was performed to examine robustness to motion in a subset of patients. Following processing, LVEF of the processed images was compared to unprocessed images.
Results
The phantom study indicated that volume measurements could be performed at a DLP of 50 mGycm, 90 kVp, 1 mm slice thickness and 256x256 matrix. The simulation study suggested the dose could be reduced to a mean simulated CTDIvol of 2.5 mGy at approximately 5\% of initial dose.
In the clinical study, CT-derived absolute LVEF showed positive bias compared to MRI, from 2 to 10\%. CT classified patients with AUCs of above 0.99. For CAP CT, the proportion of diagnostic scans did not differ from previous scans. Mean DLP for LVEF CT was 57.1 mGycm, for an effective dose of 1.34 mSv, compared to 5.68 mSv for MUGA. Reader agreement was excellent for MRI and CT but moderate for MUGA.
Simulated reduction of temporal resolution did not affect CT LVEF measurements in the threshold-based mode. Respiratory artefacts did not produce systematic bias but affected limits of agreement
Conclusion
CT can reliably classify reduced LVEF below 50\% at one-quarter the dose of MUGA but with significant absolute LVEF bias. Measurements were robust to reduced temporal resolution and simulated artefacts. A one-stop CT LVEF and CAP protocol was feasible. Results should be validated in serial follow-up studies.
CT can accurately measure left ventricular ejection fraction (LVEF) but with a high radiation dose. The potential for dose reduction in CT LVEF was explored and results used in designing a protocol to be validated clinically. A one-stop LVEF and chest-abdomen-pelvis (CAP) CT protocol was tested, for use in, e.g., in monitoring for chemotherapy-induced cardiotoxicity to potentially replace nuclear medicine MUGA scans.
Methods
Exploratory studies were performed to determine a lower radiation dose limit using 3D-printed heart phantoms and simulated dose reduction in clinical images. The phantoms were produced using 3D-printing of a clinical cardiac CT in both sys- and diastolic phases, and were scanned at multiple parameter combinations in an anthropormorphic chest phantom.
A series of clinical, functional cardiac CT scans were subjected to simulated noise reduction at 25, 10, 5 and 2 \% of the originally applied dose, using a custom-made algorithm. Following validation, LV volumetric measurements were performed at all dose levels and compared to full-dose scans with Bland-Altman plots.
From these results, a clinical protocol was designed.
The protocol was tested on 82 patients, with cardiac MRI as a reference. The protocol was combined with CAP CT with an adapted injection protocol in a subgroup of patients. 49 patients also had MUGA scans performed. Absolute agreement and classification accuracy for reduced LVEF<50\% were examined with Bland-Altman plots and ROC. CAP image quality was compared with previous scans using Visual Grading Characteristics. Inter-reader agreement was measured with ICC.
Finally, simulation of reduced temporal resolution and discontinuity artefacts was performed to examine robustness to motion in a subset of patients. Following processing, LVEF of the processed images was compared to unprocessed images.
Results
The phantom study indicated that volume measurements could be performed at a DLP of 50 mGycm, 90 kVp, 1 mm slice thickness and 256x256 matrix. The simulation study suggested the dose could be reduced to a mean simulated CTDIvol of 2.5 mGy at approximately 5\% of initial dose.
In the clinical study, CT-derived absolute LVEF showed positive bias compared to MRI, from 2 to 10\%. CT classified patients with AUCs of above 0.99. For CAP CT, the proportion of diagnostic scans did not differ from previous scans. Mean DLP for LVEF CT was 57.1 mGycm, for an effective dose of 1.34 mSv, compared to 5.68 mSv for MUGA. Reader agreement was excellent for MRI and CT but moderate for MUGA.
Simulated reduction of temporal resolution did not affect CT LVEF measurements in the threshold-based mode. Respiratory artefacts did not produce systematic bias but affected limits of agreement
Conclusion
CT can reliably classify reduced LVEF below 50\% at one-quarter the dose of MUGA but with significant absolute LVEF bias. Measurements were robust to reduced temporal resolution and simulated artefacts. A one-stop CT LVEF and CAP protocol was feasible. Results should be validated in serial follow-up studies.
| Original language | English |
|---|---|
| Awarding Institution |
|
| Supervisors/Advisors |
|
| Date of defence | 13. Dec 2023 |
| Publisher | |
| Publication status | Published - Apr 2024 |
| Externally published | Yes |