RESUMO
BACKGROUND AND PURPOSE: Accurate classification of ischemic stroke subtype is important for effective secondary prevention of stroke. We used diffusion-weighted image (DWI) and atrial fibrillation (AF) data to train a deep learning algorithm to classify stroke subtype. METHODS: Model development was done in 2,988 patients with ischemic stroke from three centers by using U-net for infarct segmentation and EfficientNetV2 for subtype classification. Experienced neurologists (n=5) determined subtypes for external test datasets, while establishing a consensus for clinical trial datasets. Automatically segmented infarcts were fed into the model (DWI-only algorithm). Subsequently, another model was trained, with AF included as a categorical variable (DWI+AF algorithm). These models were tested: (1) internally against the opinion of the labeling experts, (2) against fresh external DWI data, and (3) against clinical trial dataset. RESULTS: In the training-and-validation datasets, the mean (±standard deviation) age was 68.0±12.5 (61.1% male). In internal testing, compared with the experts, the DWI-only and the DWI+AF algorithms respectively achieved moderate (65.3%) and near-strong (79.1%) agreement. In external testing, both algorithms again showed good agreements (59.3%-60.7% and 73.7%-74.0%, respectively). In the clinical trial dataset, compared with the expert consensus, percentage agreements and Cohen's kappa were respectively 58.1% and 0.34 for the DWI-only vs. 72.9% and 0.57 for the DWI+AF algorithms. The corresponding values between experts were comparable (76.0% and 0.61) to the DWI+AF algorithm. CONCLUSION: Our model trained on a large dataset of DWI (both with or without AF information) was able to classify ischemic stroke subtypes comparable to a consensus of stroke experts.
RESUMO
OBJECTIVES: Differentiation among benign salivary gland tumours, Warthin tumours (WTs), and malignant salivary gland tumours is crucial to treatment planning and predicting patient prognosis. However, differentiation of those tumours using imaging findings remains difficult. This study evaluated the usefulness of elasticity determined from diffusion-weighted image (DWI)-based virtual MR elastography (MRE) compared with conventional magnetic resonance imaging (MRI) findings in differentiating the tumours. METHODS: This study included 17 benign salivary gland tumours, 6 WTs, and 11 malignant salivary gland tumours scanned on neck MRI. The long and short diameters, T1 and T2 signal intensities, tumour margins, apparent diffusion coefficient (ADC) values, and elasticity from DWI-based virtual MRE of the tumours were evaluated. The interobserver agreement in measuring tumour elasticity and the receiver operating characteristic (ROC) curves were also assessed. RESULTS: The long and short diameters and the T1 and T2 signal intensities showed no significant difference among the 3 tumour groups. Tumour margins and the mean ADC values showed significant differences among some tumour groups. The elasticity from virtual MRE showed significant differences among all 3 tumour groups and the interobserver agreement was excellent. The area under the ROC curves of the elasticity were higher than those of tumour margins and mean ADC values. CONCLUSION: Elasticity values based on DWI-based virtual MRE of benign salivary gland tumours, WTs, and malignant salivary gland tumours were significantly different. The elasticity of WTs was the highest and that of benign tumours was the lowest. The elasticity from DWI-based virtual MRE may aid in the differential diagnosis of salivary gland tumours.