Renal MRI radiomics in Beckwith-Wiedemann syndrome: a novel imaging approach for genotype identification

Abstract

Purpose: To valuate the role of nonmalignant nephrological findings and renal MRI radiomics in differentiating molecular subtypes of Beckwith-Wiedemann syndrome (BWS).

Materials and methods: Clinical data and abdominal MRI scans of 49 patients who underwent partial glossectomy between July 2019 and March 2024 were retrospectively analysed. Patients were categorized into two subtypes: BWS^UPD+IC1 (24 cases, with a predisposition to renal involvement) and BWS^IC2 (25 cases, with a lower risk of renal involvement), based on genetic testing. Pearson correlation analysis was conducted to evaluate the relationship between patients' age and renal volume. Radiomic features derived from the T2WI sequence and the ADC map were selected to construct single-sequence and combined models. Delong test was used to compare the performance of the models.

Results: Clinically, the BWS^UPD+IC1 subtype exhibited a lower incidence of ear creases/pits (P = 0.048) and omphalocele/umbilical hernia (P = 0.032) compared to the BWS^IC2 subtype. Abdominal MRI findings indicated the BWS^UPD+IC1 subtype had larger total renal volume (P = 0.017) and a weaker correlation between total renal volume and patients' age (r = 0.38). Notably, 91.84% (45/49) of BWS patients exhibited a total renal volume exceeding the normal population's upper limit, with the IC1 subtype demonstrating the largest mean volume. The BWS^UPD+IC1 subtype showed higher incidences of nonmalignant renal (P = 0.013) and non-renal abdominal abnormalities. The T2WI, ADC, and combined models achieved the highest area under the receiver operating characteristic (ROC) curves (AUCs) of 0.837, 0.882 and 0.954 (P > 0.05), respectively.

Conclusion: Nonmalignant renal abnormalities and MRI radiomics models have potential as alternative imaging tools for the identification of renal predisposition genotypes and the surveillance of renal size change in BWS patients.

Keywords: Beckwith–Wiedeman syndrome; Genotype; Kidney; Radiomics.

Fig. 1

The flowchart of patient recruitment. *4 patients underwent both WES and MLPA testing, 1 patient was CDKN1C, 1 patient was negative result. WES, whole exome sequencing; CMA, chromosomal microarray analysis; MLPA, multiplex ligation-dependent probe amplification; LOH, loss of heterozygosity

Fig. 2

Radiomics workflow of this study. ROI, region of interest; VOI, voxel of interest; GLCM, gray level of co-occurrence matrix; GLRLM, gray level of run length matrix; GLSZM, gray level of size zone matrix; GLDM, gray level of dependence matrix; NGTDM, neighbouring gray tone difference matrix; GBDT, gradient boosting decision tree; KNN, k-nearest neighbors; PLSDA, partial least squares discriminant analysis; QDA, quadratic discriminant analysis; SGD, stochastic gradient descent; SVM, support vector machine

Fig. 3

The variation range of total renal volume in normal children aged 6 to 48 months (in half-year intervals), along with the scatter plot and correlation between the total renal volume and the patient’s age in BWS patients. The three dashed lines, ranging from dark to light green, represented the upper limit (mean + 2SD), the mean value, and the lower limit (mean - 2SD) respectively. To distinguish between the UPD and IC1 subtypes, the red circular points for the IC1 subtype were outlined with black circles. Four patients had the total renal volume below the normal upper limit, which were one UPD patient (10 months), one IC1 patient (16 months), and two IC2 patients (12 and 18 months). *Based on reference [17]

Fig. 4

MRI of nephromegaly and complex cysts in a 2-year 3-month-old boy with BWS^IC1 and a vascular malformation in the back. Arrows indicated an isointense signal on the fat-suppressed T2WI (a) and a hyperintense signal on the DWI (b) in the left kidney. A smaller hyperintense signal was observed in the right kidney on the DWI (c), which was not present on the T2WI

Fig. 5

MRI of a simple cortical cyst in a 2 year, 9-month-old boy with BWS^IC2. Arrows indicated a subtle hyperintense signal in the left kidney on the fat-suppressed T2WI (a). The lesion was invisible on DWI (b) and showed no diffusion restriction

Fig. 6

MRI of suspected multifocal vascular malformations in a 11-month-old girl with BWS^UPD. Arrows indicated multiple patchy hyperintense signals in the liver on the fat-suppressed T2WI (a) and DWI (b). A similar signal was observed in the left kidney on T2WI and DWI (c, d)

Fig. 7

MRI of hydronephrosis and pancreatic hypertrophy in a 11-month-old boy with BWS^UPD. The right renal hydronephrosis (arrow) and localized enlargement of the pancreatic body and tail (arrowhead) were observed on the fat-suppressed T2WI

Fig. 8

ROC curves of each fold and the best mean values for the T2WI model (a, b), ADC model (c, d) and combined model (e, f) on the training and test datasets