Evaluation of an automated intracranial aneurysm detection and rupture analysis approach using cascade detection and classification networks.

Intracranial aneurysm is commonly found in human brains especially for the elderly, and its rupture accounts for a high rate of subarachnoid hemorrhages. However, it is time-consuming and requires special expertise to pinpoint small aneurysms from computed tomography angiography (CTA) images. Deep learning-based detection has helped improve much efficiency but false-positives still render difficulty to be ruled out. To study the feasibility of deep learning algorithms for aneurysm analysis in clinical applications, this paper proposes a pipeline for aneurysm detection, segmentation, and rupture classification and validates its performance using CTA images of 1508 subjects. A cascade aneurysm detection model is employed by first using a fine-tuned feature pyramid network (FPN) for candidate detection and then applying a dual-channel ResNet aneurysm classifier to further reduce false positives. Detected aneurysms are then segmented by applying a traditional 3D V-Net to their image patches. Radiomics features of aneurysms are extracted after detection and segmentation. The machine-learning-based and deep learning-based rupture classification can be used to distinguish ruptured and un-ruptured ones. Experimental results show that the dual-channel ResNet aneurysm classifier utilizing image and vesselness information helps boost sensitivity of detection compared to single image channel input. Overall, the proposed pipeline can achieve a sensitivity of 90 % for 1 false positive per image, and 95 % for 2 false positives per image. For rupture classification the area under curve (AUC) of 0.906 can be achieved for the testing dataset. The results suggest feasibility of the pipeline for potential clinical use to assist radiologists in aneurysm detection and classification of ruptured and un-ruptured aneurysms.

Ischemia in intracerebral hemorrhage: A comparative study of small-vessel and large-vessel diseases.

OBJECTIVE: This study aimed to compare effects of cerebral small-vessel disease (cSVD) burden and cerebral artery stenosis (CAS) on acute ischemia in intracerebral hemorrhage (ICH) and their interaction with mean arterial pressure (MAP) change. METHODS: We recruited consecutive patients with acute primary ICH. Brain magnetic resonance imaging and angiography were performed to quantify diffusion-weighted imaging (DWI) lesions, CAS, and cSVD markers, which were calculated for the total cSVD score. Multivariable regression models were adopted to explore their associations by DWI lesions size (<15 vs. ≥15 mm) and median MAP change stratification. RESULTS: Of 305 included patients (mean age 59.5 years, 67.9% males), 77 (25.2%) had DWI lesions (small, 79.2%; large, 20.8%) and 67 (22.0%) had moderate and severe CAS. In multivariable analysis, small DWI lesions were independently associated with higher total cSVD score (odds ratio [OR] 1.81, 95% confidence interval [CI] 1.36-2.41). and large DWI lesions were associated with more severe CAS (OR 2.51, 95% CI 1.17-5.38). This association was modified by MAP change (interaction p = 0.016), with stratified analysis showing an increased risk of large DWI lesions in severe CAS with greater MAP change (≥44 mmHg) (OR 3.48, 95% CI 1.13-10.74) but not with mild MAP change (<44 mmHg) (OR 1.21, 95% CI 0.20-7.34). INTERPRETATION: Total cSVD burden is associated with small DWI lesions, whereas the degree of CAS is associated with large DWI lesions, specifically with greater MAP change, suggesting that large-artery atherosclerosis may be involved in ischemic brain injury, which is different from small-vessel pathogenesis in ICH.

Confirmation of the abnormal lipid metabolism as a risk factor for the disease of leukoaraiosis.

Our purpose is to screen out medical history indicators and test indicators linked to lipid metabolism which is closely correlated to leukoaraiosis (LA), and to build assistant diagnosis model based on support vector machine (SVM), which provided theoretical evidence for genesis and development of LA. One thousand LA patients who underwent magnetic resonance imaging (MRI) examination in Imaging Department was retrospectively analyzed and divided into LA group and non-LA group in accordance with examination results. Detailed clinical statistics of the two groups were collected, including test indicators related to lipid metabolism, such as total cholesterol (TC), triglyceride (TG), low density lipoprotein (LDL), high density lipoprotein (HDL), medical history indicators, age, sex, diabetes, hypertension, hyperlipidemia, history of intracranial infection, history of cerebral hemorrhage, cerebral infarction, lacunar infarction and relevant biochemical indexes. The study shows that patients' incidence of LA was 31.10%; in accordance with Logistic analysis, the incidence of LA is significantly correlated to factors like age, hypertension, history of cerebral hemorrhage, cerebral infarction, lacunar infarction and triglyceride elevation; two SVMs, one including all variables and the other containing all screened variables were successfully established, and the former's accuracy, specificity and sensitivity respectively were 85.0%, 85.0% and 85.0% while the latter's 90.0%, 100.0% and 80.0%. Test indicators and medical history indicators of lipid metabolism correlated to LA were screened out successfully. Meanwhile, an effective SVM model also was built successfully, which is able to predict LA relatively accurately and can be used as assistant diagnostic tool for clinicians.

[The role of serum hepcidin and ferroportin1 in placenta on iron transfer from mother to fetus].

OBJECTIVE: To detect the concentration of serum hepcidin and the mRNA expression level of ferroportin1 (FPN1) in the placenta membrane from full term pregnant women with different degree of iron deficiency, and explore their roles for iron transport in placental. METHODS: The concentration of HGB, serum iron (SI) and serum ferritin (SF) of mothers and infants were detected in 55 full term pregnant women and neonates. The expression level of FPN1 mRNA in placental was detected by the RT-PCR technique. The concentration of serum hepcidin was detected by double antibody sandwich biotin avidin enzyme-linked immunosorbent assay. The serum hepcidin level and the FPN1 mRNA expression in the full term placenta from different maternal iron status were compared in three groups. RESULTS: There were no significant differences in the cord blood HGB, SI and SF of newborns from pregnant women with different iron status (P>0.05). The concentration of serum hepcidin of pregnant women among normal, iron deficiency and mild iron deficiency anemia were (193.637±52.219), (176.523±43.875), and (147.623±37.768) µg/L respectively, with statistical significance (F=3.872, P=0.027). The expression levels of FPN1 mRNA among three groups were 0.462±0.077, 0.507±0.074 and 0.551±0.104 respectively, with statistical significance (F=4.767, P=0.013). A negative correlation between maternal serum hepcidin and placental FPN1 mRNA (r=-0.383, P=0.004) was identified. CONCLUSION: There were no significant differences in the iron status of corresponding newborns from pregnant women with different iron status. With the severity of maternal iron deficiency, the concentration of serum hepcidin was down-regulated, while the expression of FPN1 mRNA in placenta was up-regulated.