The Use of a Software-Assisted Method to Estimate Fetal Weight at and Near Term Using Magnetic Resonance Imaging

Abstract

Objective: The aim of this study was to apply a semi-automated calculation method of fetal body volume and, thus, of magnetic resonance-estimated fetal weight (MR-EFW) prior to planned delivery and to evaluate whether the technique of measurement could be simplified while remaining accurate. Methods: MR-EFW was calculated using a semi-automated method at 38.6 weeks of gestation in 36 patients and compared to the picture archiving and communication system (PACS). Per patient, 8 sequences were acquired with a slice thickness of 4-8 mm and an intersection gap of 0, 4, 8, 12, 16, or 20 mm. The median absolute relative errors for MR-EFW and the time of planimetric measurements were calculated for all 8 sequences and for each method (assisted vs. PACS), and the difference between the methods was calculated. Results: The median delivery weight was 3,280 g. The overall median relative error for all 288 MR-EFW calculations was 2.4% using the semi-automated method and 2.2% for the PACS method. Measurements did not differ between the 8 sequences using the assisted method (p = 0.313) or the PACS (p = 0.118), while the time of planimetric measurement decreased significantly with a larger gap (p < 0.001) and in the assisted method compared to the PACS method (p < 0.01). Conclusion: Our simplified MR-EFW measurement showed a dramatic decrease in time of planimetric measurement without a decrease in the accuracy of weight estimates.

Prenatal prediction of postnatal large‐for‐dates neonates using a simplified MRI method: comparison with conventional 2D ultrasound estimates

“…This was a single‐institution cross‐sectional study conducted at the Department of Radiology and the Fetal Medicine Unit of the University Hospital Brugmann, Université Libre de Bruxelles, Brussels, Belgium. Following the recent development by our group of purpose‐designed software to assist planimetric measurement of FBV, we retrieved retrospectively from our database data of women satisfying the entry criteria for the present study and applied the new method of MRI‐EFW.…”

“…FBV was delineated using a purpose‐designed semi‐automated computer‐assisted method via a home‐built user interface programmed with the MATLAB (matrix laboratory) environment (Matlab 2013a, The MathWorks, Inc, Natick, MA, USA) as previously described (Figure ). Semi‐automated FBV measurements were performed by a single maternal–fetal medicine specialist (C.K.)…”

Protocol for the prospective observational clinical study: estimation of fetal weight by MRI to PREdict neonatal MACROsomia (PREMACRO study) and small-for-gestational age neonates

“…A more accurate method of EFW could produce better outcomes by restricting interventions such as induction of labour and caesarean section to those fetuses at greatest risk of suspected macrosomia. Some publications have already demonstrated that MRI-derived EFW close to delivery is more accurate than US, with a precision superior to that of US using birth weight as a golden standard,14–23 and a recent meta-analysis has confirmed this promising accuracy 24…”

The use of magnetic resonance imaging in the prediction of birthweight

“…In the years that followed from 2013 to 2018, our group has continued to publish in this area, further validating, refining, and testing the method in different situations and types of pregnancies. The first advance was the progression from a manual technique for the assessment of FBV to a computer‐aided semiautomatic technique, which reduced the measurements times from 30 minutes down to 5 minutes but without compromising accuracy . We also showed that the number of MRI images used for the planimetric technique did not significantly affect the fetal weight estimations and that the semiautomatic technique for FBV is highly reproducible with interobserver and intraobserver agreement and intraclass coefficients (ICCs) of 0.99 and 0.98, respectively .…”

“…Specific software to facilitate the fetal semisegmentation are now available from various imaging manufacturers, but we developed our own user interface programmed with a MATLAB (matrix laboratory) environment (MATLAB 2013a, the MathWorks, Inc, Natick, MA, US) to help delineate FBV. The technique has been described previously, but essentially, the program performs an automatic search to identify voxels that represent the fetal skin. The operator selects a point on the fetal border and the program then automatically searches for a path of skin pixels that connects the selected points, in addition to adding new points.…”