Discussion
This study demonstrates that MRI of the fetal thorax can be processed by 3D DSVR reconstruction, and subsequent lung volumetry can be performed with similar volumes generated to the 2D-manual segmentation derived volumes currently used in clinical practice for O/E TLV assessment. Furthermore, 3D lung segmentations potentially provide more accurate TLV estimation since they minimise the errors on the conventional 2D-based approach affected by the combination of misalignment of slices and interpolation. The analysis of the quantitative comparison results as well as the general visual inspection of 2D vs. 3D segmentations suggest that using 3D DSVR-based TLV assessment may potentially resolve one of the current limitations to the validity of MRI derived volumes in predicting outcomes as reliably as LHR13. This series demonstrates that DSVR method provides the means for accurate calculation of lung volumes for both normal and abnormal cases. This is especially useful for CDH cases for more accurate evaluation of the observed to expected lung volume ratios. The recent advances in image processing methods reduce variability in segmentation14, further improving the reliability of the volumetry results. We anticipate that MR based lung volumetry may soon be validated as a reliable and more accurate prognostic indicator than the current use of the LHR; since it would invariably more closely predict true TLV. Furthermore, the ability to outline lesion volumes in cases of CLL will likely produce a prognostic marker more faithful to true lesion size than the currently utilised CVR where cross sectional areas are utilised to estimate lesion volume relative to fetal size. The consistency of such a method will certainly deliver on the recent plea for consistency in prognostication of these cases9. It should be noted that although we have provided lesion volume proportionate to TLV, it would be possible to segment any aspect of the fetus and derive a corresponding volume ratio proportional to this as well as the relative position of the organs (e.g., the degree of liver herniation or position of the stomach in CDH cases)
Recently, it has also been demonstrated that DSVR reconstructed MRI images may provide additional diagnostic information for complex fetal anomalies1. The illustrative cases included within this dataset also show that DSVR will produce datasets that are more accessible to the clinician by resolving motion corruption with a resolution superior to that of the original input stacks. This will certainly aid in prenatal counselling and planning of the operative approach in neonatal thoracic surgery; the reliable depiction of systemic feeding vessels in BPS, combined with US Doppler assessment may reduce the reliance on early postnatal CT imaging 15.
Regarding longitudinal assessment of CDH or CLL, it should be recognised that the repeated use of MRI is considerably more expensive, time consuming and lengthy compared to an expert-performed ultrasound. Therefore, we anticipate that there will continue to be a role for ultrasound in the monitoring of such conditions to delivery. Indeed, serial CVR assessment has recently been shown to be effective in predicting a need for perinatal intervention in cases of CLL16.