Automatic Facial Recognition of Williams-Beuren Syndrome Based on Deep Convolutional Neural Networks.

Background: Williams-Beuren syndrome (WBS) is a rare genetic syndrome with a characteristic "elfin" facial gestalt. The "elfin" facial characteristics include a broad forehead, periorbital puffiness, flat nasal bridge, short upturned nose, wide mouth, thick lips, and pointed chin. Recently, deep convolutional neural networks (CNNs) have been successfully applied to facial recognition for diagnosing genetic syndromes. However, there is little research on WBS facial recognition using deep CNNs. Objective: The purpose of this study was to construct an automatic facial recognition model for WBS diagnosis based on deep CNNs. Methods: The study enrolled 104 WBS children, 91 cases with other genetic syndromes, and 145 healthy children. The photo dataset used only one frontal facial photo from each participant. Five face recognition frameworks for WBS were constructed by adopting the VGG-16, VGG-19, ResNet-18, ResNet-34, and MobileNet-V2 architectures, respectively. ImageNet transfer learning was used to avoid over-fitting. The classification performance of the facial recognition models was assessed by five-fold cross validation, and comparison with human experts was performed. Results: The five face recognition frameworks for WBS were constructed. The VGG-19 model achieved the best performance. The accuracy, precision, recall, F1 score, and area under curve (AUC) of the VGG-19 model were 92.7 ± 1.3%, 94.0 ± 5.6%, 81.7 ± 3.6%, 87.2 ± 2.0%, and 89.6 ± 1.3%, respectively. The highest accuracy, precision, recall, F1 score, and AUC of human experts were 82.1, 65.9, 85.6, 74.5, and 83.0%, respectively. The AUCs of each human expert were inferior to the AUCs of the VGG-16 (88.6 ± 3.5%), VGG-19 (89.6 ± 1.3%), ResNet-18 (83.6 ± 8.2%), and ResNet-34 (86.3 ± 4.9%) models. Conclusions: This study highlighted the possibility of using deep CNNs for diagnosing WBS in clinical practice. The facial recognition framework based on VGG-19 could play a prominent role in WBS diagnosis. Transfer learning technology can help to construct facial recognition models of genetic syndromes with small-scale datasets.

Adding DSC PWI and DWI to BT-RADS can help identify postoperative recurrence in patients with high-grade gliomas.

BACKGROUND: The Brain Tumor Reporting and Data System (BT-RADS) category 3 is suitable for identifying cases with intermediate probability of tumor recurrence that do not meet the Response Assessment in Neuro-Oncology (RANO) criteria for progression. The aim of this study was to evaluate the added value of dynamic susceptibility contrast-enhanced perfusion-weighted imaging (DSC PWI) and diffusion-weighted imaging (DWI) to BT-RADS for differentiating tumor recurrence from non-recurrence in postoperative high-grade glioma (HGG) patients with category 3 lesions. METHODS: Patients with BT-RADS category 3 lesions were included. The maximal relative cerebral blood volume (rCBVmax) and the mean apparent diffusion coefficient (ADCmean) values were measured. The added value of DSC PWI and DWI to BT-RADS was evaluated by receiver operating characteristic (ROC) curve analysis. RESULTS: Fifty-one of 91 patients had tumor recurrence, and 40 patients did not. There were significant differences in rCBVmax and ADCmean between the tumor recurrence group and non-recurrence group. Compared to BT-RADS alone, the addition of DSC PWI to BT-RADS increased the area under curve (AUC) from 0.76 (95% confidence interval [CI] 0.66-0.84) to 0.90 (95% CI 0.81-0.95) for differentiating tumor recurrence from non-recurrence. The addition of DWI to BT-RADS increased the AUC from 0.76 (95% CI 0.66-0.84) to 0.88 (95% CI 0.80-0.94). The combination of BT-RADS, DSC PWI, and DWI exhibited the best diagnostic performance (AUC = 0.95; 95% CI 0.88-0.98) for differentiating tumor recurrence from non-recurrence. CONCLUSION: Adding DSC PWI and DWI to BT-RADS can significantly improve the diagnostic performance for differentiating tumor recurrence from non-recurrence in BT-RADS category 3 lesions.

Ultra-low-dose multiphase CT angiography derived from CT perfusion data in patients with middle cerebral artery stenosis.

PURPOSE: Computed tomography (CT) perfusion (CTP) source images contain both brain perfusion and cerebrovascular information, and may allow a dynamic assessment of collaterals. The purpose of the study was to compare the image quality and the collaterals identified on multiphase CT angiography (CTA) derived from CTP datasets (hereafter called CTPA) reconstructed with iterative model reconstruction (IMR) algorithm in patients with middle cerebral artery (MCA) steno-occlusion with those of routine CTA. METHODS: Consecutive patients with a unilateral MCA steno-occlusion underwent non-contrast CT (NCCT), CTP, and CTA. CTPA images were reconstructed from CTP datasets. The vascular attenuation, image noise, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) of routine CTA and CTPA were measured and analyzed by Student's t test. Subjective image quality and collaterals were scored and compared using the Wilcoxon signed-rank test. RESULTS: Fifty-eight patients (mean age 61.7 years, 78% males, median National Institutes of Health Stroke Scale score = 12) were included. The effective radiation dose of CTP was 1.28 mSv. The vascular attenuation, SNR, CNR, and the image quality of CTPA were considerably higher than that of CTA (all, p < 0.001). Collaterals were rated higher on CTPA compared with CTA (1.79 ± 0.64 vs. 1.22 ± 0.84, p < 0.001). Fifty-three percent of patients with poor collaterals assessed on single-phase CTA had good collaterals on CTPA. CONCLUSION: CTPA derived from CTP datasets reconstructed with IMR algorithm offers image quality comparable to routine CTA and provides time-resolved evaluation of collaterals in patients with MCA ischemic disease.