The feasibility of using PDW MRI to assist tandem delineation for cervical cancer patients treated with HDR brachytherapy was proved in this work. It was shown that PDW MRI had a better uterus-tandem contrast compared to T2W MRI. Qualitatively, the tandem had a better and more uniform shape on PDW MRI. Quantitatively, the difference between measured and physical diameters of the tandem was significantly reduced by using PDW MRI, confirming PDW MRI had reduced artifacts.
Although one of the standard methods to reduce geometric distortion and signal dropout is to increase the bandwidth and/or matrix size, it may not be a practical solution to improve image quality of T2W MRI and reduce effort in tandem delineation. In T2W MRI, the uterus signal is typically very low because of the use of long TE. The uterus-tandem contrast is barely enough for the delineation purpose. Increasing the bandwidth and/or matrix size, however, will significantly reduce the image SNR and thus the uterus-tandem contrast, making tandem delineation extremely difficult, if not impossible at all. PDW MRI uses a very short (5.5 ms vs. 90 ms). Therefore, the uterus signal is much higher in PDW MRI compared to T2W MRI, which not only increases the uterus-tandem contrast, but also allows us to use thinner slices to further reduce the geometric distortion and signal dropout around the titanium tandem. T1W MRI utilizes a TE which is comparable to PDW MRI. When T1W and PDW images are acquired using the same TE, bandwidth and spatial resolution, the two image datasets should exhibit similar amount of metal artifacts. On the other hand, T1W MRI utilizes a relatively shorter TR compared to PDW MRI, which makes it less SNR efficient. The long TR used in PDW MRI allows the magnetization to be fully relaxed before the next excitation. Therefore, more magnetization will be available at the beginning of each RF excitation, resulting in higher signal (or SNR) in PDW MRI. The SNR advantage of PDW MRI may be traded off for better spatial resolution, which will further reduce metal artifacts.
Image fusion between the T2W and PDW images was accomplished by using the coordinates stored in the DICOM header. This was warranted in this study because 1) the tabletop supporting patients did not move between the T2W and PDW MRI scans, all MRI images were acquired by changing the imaging hardware such as gradients; and 2) patients were still under medication for pain control which made them sleepy and less likely to move. None of the 10 patients included in this study showed significant movement between T2W and PDW MRI. If in future we do observe significant movement between scans, we can implement sophisticated fusion techniques based on patient’s anatomy to ensure the accuracy of the quantitative analysis.
The measured diameters from MRI images had a relatively large variance among patients. The largest measurement was 0.79 cm in one of the T2W images whereas the smallest measurement was 0.29 cm in one of the PDW images. The standard deviations in diameter measurement were 0.15 cm and 0.11 cm for T2W and PDW images, respectively. After examining Table 1, we noticed that there was a good correlation between the measured diameters from T2W images and PDW images. Specifically, if the measured diameter from the T2W image was large, the corresponding measured tandem diameter from the PDW image was also large and vice versa. In addition, we found that the measured diameter of the tandem correlated well with the curvature of the tandem. Patients using a tandem with bigger curvature tended to have a larger measured diameter. If patients were grouped by the curvature of the tandem, 3 patients had tandem curvature of 30° and 7 patients had tandem curvature of 45°. For the group with tandem curvature of 30°, the standard deviation of measured diameters reduced to 0.05 cm by using the T2W images and to 0.04 cm by using the PDW images. Similarly, for the group with tandem curvature of 45°, the standard deviation of measured diameters reduced to 0.10 cm by using the T2W images and to 0.08 cm by using the PDW images. We also noticed that some of the measured diameters were smaller than the physical diameter of the tandem. This may be explained by the relative location of the slices to the tandem. If the tandem is equally distributed into two adjacent para-sagittal slices along its axis, then each slice contains half of the tandem. Since the radius of the tandem is smaller than the slice thickness, for any voxels within the region of the tandem, it contains both tandem and the uterus. This partial volume effect may lower the contrast between the tandem and the uterus, making the measured diameter of the tandem smaller than its physical diameter. All these observations suggest that the measured diameter of the tandem is related to the curvature of the tandem, the location of slices, and the placement of the receiving coil and the applicator as well. Further investigation is needed and will be conducted in the future to fully understand the causes.
The focus of this retrospective study was to show the feasibility of using PDW MRI as an option to assist tandem delineation for cervical cancer patients treated with HDR brachytherapy. Other options to assist tandem delineation include acquiring additional x-ray images, CT images, T1W MRI or other MRI image-sets. Evaluating those options to find the best choice is a very interesting topic and of great clinical importance. However, it is out of the scope of this study. Such kind of study will be planned and performed in the future after being carefully designed to minimize the bias in the process of comparing images from different imaging modalities. Another interesting project is to compare dwell positions determined by MR images (T2W, T1W and PDW) and CT images. In our preliminary study on phantom , we found that mean differences between activated dwells on MRI vs. CT were approximately 0.1cm or less, 0.2 cm and 0.1 cm in RL (right-left), FH (foot-head), and AP (anterior-posterior) directions, respectively. Comparisons between T2W and PDW MRI showed about 0.1 cm difference in the FH direction and negligible differences in the other directions. Comparison of dwell positions using patient image datasets is currently under investigation.