MRI scout images can detect the acute intracerebral hemorrhage on CT

Identifying Cerebral Large Vessel Occlusion in Acute Ischemic Stroke by MRI Positioning Scanning

Various approaches have been tried for acute ischemic stroke (AIS) treatment to shorten the time from onset to recanalization. MRI positioning scanning (PS), which must be taken before any MRI sequences, was examined whether it can detect cerebral large vessel occlusion. A total of 68 consecutive patients with AIS who underwent MRI and were treated with intravenous recombinant tissue plasminogen activator or mechanical thrombectomy at our hospital were retrospectively included in this study. Occluded vessels were identified on the axial or coronal views of PS images, and these images were compared with 3D time-of-flight MRA and digital subtraction angiogram. The sensitivities, positive predictive values (PPVs), and negative predictive values (NPVs) for internal carotid artery (ICA), the proximal M1, distal M1, and M2 segment of the middle cerebral artery occlusion were assessed, and the number of PS slices was assessed. The sensitivities of the axial slices for ICA, proximal M1, distal M1, and M2 occlusion were 62%, 21%, 35%, and 86%, respectively. The PPVs of the axial slices for ICA, proximal M1, distal M1, and M2 occlusion were 81%, 88%, 100%, and 97%, respectively, and the NPVs of the axial slices for ICA, proximal M1, distal M1, and M2 occlusion were 94%, 90%, 86%, and 100%, respectively. The detection rate for the ICA was significantly higher with three axial slices (91%) than with two slices (47%) (p <0.01). MRI PS is warranted to be referred to detect large cerebral vessel occlusion.

Readability of extraspinal organs on scout images of lumbar spine MRI according to different protocols

Background

Scout images of lumbar spine MRI often include the extraspinal organs, which are barely included in routine MRI and can be a potential cause of lumbar pain.

Purpose

To evaluate the readability of scout images for extraspinal organs in lumbar spine MRI according to different protocols.

Materials and methods

A total of 150 patients who underwent 1.5 T or 3 T lumbar spine MRI from March to September 2015 at three hospitals with different scout image protocols, were selected. Two radiologists independently reviewed the scout images to investigate whether exclusive diagnosis of major diseases involving the femoral head, femoral neck, sacroiliac joint, and kidneys was possible. Readability levels were divided into four categories: definitely, possibly, limited, and non-evaluable. The readability of scout images according to the protocols was compared using Chi-square test. Interobserver agreement for the readability level of scout images was assessed using weighted κ statistics.

Results

Of 150 patients, “definitely evaluable” cases classified by two readers were 50–62 (33.3–41.3%) for femoral head (κ = 0.63–0.71), 37–66 (24.7–44.0%) for femoral neck (κ = 0.41–0.48), 72–93 (48.0–62.0%) for sacroiliac joint (κ = 0.35–0.37), and 63–73 (42.0–48.7%) for kidneys (κ = 0.45–0.47). More than 50% of femoral heads were classified as readable (definitely or possible evaluable) cases by two readers with excellent interobserver agreement. The readability level of scout images was significantly different according to image protocols including the MRI sequence, number of coronal plane slices, and intersection gap of coronal plane slices (p≤0.015).

Conclusion

Scout images of lumbar spine MRI may be readable enough to rule out some major diseases of extraspinal organs. Standardization of the protocol will be needed to validate the potential role of scout images for screening extraspinal organs.

Multimodal 3D DenseNet for IDH Genotype Prediction in Gliomas

Non-invasive prediction of isocitrate dehydrogenase (IDH) genotype plays an important role in tumor glioma diagnosis and prognosis. Recently, research has shown that radiology images can be a potential tool for genotype prediction, and fusion of multi-modality data by deep learning methods can further provide complementary information to enhance prediction accuracy. However, it still does not have an effective deep learning architecture to predict IDH genotype with three-dimensional (3D) multimodal medical images. In this paper, we proposed a novel multimodal 3D DenseNet (M3D-DenseNet) model to predict IDH genotypes with multimodal magnetic resonance imaging (MRI) data. To evaluate its performance, we conducted experiments on the BRATS-2017 and The Cancer Genome Atlas breast invasive carcinoma (TCGA-BRCA) dataset to get image data as input and gene mutation information as the target, respectively. We achieved 84.6% accuracy (area under the curve (AUC) = 85.7%) on the validation dataset. To evaluate its generalizability, we applied transfer learning techniques to predict World Health Organization (WHO) grade status, which also achieved a high accuracy of 91.4% (AUC = 94.8%) on validation dataset. With the properties of automatic feature extraction, and effective and high generalizability, M3D-DenseNet can serve as a useful method for other multimodal radiogenomics problems and has the potential to be applied in clinical decision making.