Support Vector Machine-based Spontaneous Intracranial Hypotension Detection on Brain MRI.

BACKGROUND AND PURPOSE: To develop a fully automatic algorithm for the magnetic resonance imaging (MRI) identification of patients with spontaneous intracranial hypotension (SIH). MATERIAL AND METHODS: A support vector machine (SVM) was trained with structured reports of 140 patients with clinically suspected SIH. Venous sinuses and basal cisterns were segmented on contrast-enhanced T1-weighted MPRAGE (Magnetization Prepared-Rapid Gradient Echo) sequences using a convolutional neural network (CNN). For the segmented sinuses and cisterns, 56 radiomic features were extracted, which served as input data for the SVM. The algorithm was validated with an independent cohort of 34 patients with proven cerebrospinal fluid (CSF) leaks and 27 patients who had MPRAGE scans for unrelated reasons. RESULTS: The venous sinuses and the suprasellar cistern had the best discriminative power to separate SIH and non-SIH patients. On a combined score with 2 points, mean SVM score was 1.41 (±0.60) for the SIH and 0.30 (±0.53) for the non-SIH patients (p < 0.001). Area under the curve (AUC) was 0.91. CONCLUSION: A fully automatic algorithm analyzing a single MRI sequence separates SIH and non-SIH patients with a high diagnostic accuracy. It may help to consider the need of invasive diagnostics and transfer to a SIH center.

Spontaneous intracranial hypotension: diagnostic and therapeutic workup.

Spontaneous intracranial hypotension (SIH) is an orthostatic headache syndrome with typical MRI findings among which engorgement of the venous sinuses, pachymeningeal enhancement, and effacement of the suprasellar cistern have the highest diagnostic sensitivity. SIH is in almost all cases caused by spinal CSF leaks. Spinal MRI scans showing so-called spinal longitudinal extradural fluid (SLEC) are suggestive of ventral dural tears (type 1 leak) which are located with prone dynamic (digital subtraction) myelography. As around half of the ventral dural tears are located in the upper thoracic spine, additional prone dynamic CT myelography is often needed. Leaking nerve root sleeves typically associated with meningeal diverticulae (type 2 leaks) and CSF-venous fistulas (type 3 leaks) are proven via lateral decubitus dynamic digital subtraction or CT myelography: type 2 leaks are SLEC-positive if the tear is proximal and SLEC-negative if it is distal, and type 3 leaks are always SLEC-negative. Although 30-70% of SIH patients show marked improvement following epidural blood patches applied via various techniques definite cure mostly requires surgical closure of ventral dural tears and surgical ligations of leaking nerve root sleeves associated with meningeal diverticulae or CSF-venous fistulas. For the latter, transvenous embolization with liquid embolic agents via the azygos vein system is a novel and valuable therapeutic alternative.

High-resolution Compressed-sensing T1 Black-blood MRI : A New Multipurpose Sequence in Vascular Neuroimaging?

BACKGROUND AND PURPOSE: In vasculopathies of the central nervous system, reliable and timely diagnosis is important against the background of significant morbidity and sequelae in cases of incorrect diagnosis or delayed treatment. Magnetic resonance imaging (MRI) plays a major role in the detection and monitoring of intracranial and extracranial vascular pathologies of different etiologies, in particular for evaluation of the vessel wall in addition to luminal information, thus allowing differentiation between various vasculopathies. Compressed-sensing black-blood MRI combines high image quality with relatively short acquisition time and offers promising potential in the context of neurovascular vessel wall imaging in clinical routine. This case review gives an overview of its application in the diagnosis of various intracranial and extracranial entities. METHODS: An optimized high-resolution compressed-sensing black-blood 3D T1-weighted fast (turbo) spin echo technique (T1 CS-SPACE prototype) precontrast and postcontrast application at 3T was used for the evaluation of various vascular conditions in neuroradiology. RESULTS: In this article seven cases of intracranial and extracranial arterial and venous vasculopathies with representative imaging findings in high-resolution compressed-sensing black-blood MRI are presented. CONCLUSION: High-resolution 3D T1 CS-SPACE black-blood MRI is capable of imaging various vascular entities in high detail with whole head coverage and low susceptibility for motion artifacts and within acceptable scan times. It represents a highly versatile, non-invasive technique for the visualization and differentiation of a wide variety of neurovascular arterial and venous disorders.

Spontaneous Intracranial Hypotension.

BACKGROUND: Spontaneous intracranial hypotension (SIH) is an underdiagnosed disease. Its incidence is estimated at 5 per 100 000 persons per year. METHODS: This review is based on a selective literature search in PubMed covering the years 2000-2019, as well as on the authors' personal experience. RESULTS: The diagnostic and therapeutic methods discussed here are supported by level 4 evidence. SIH is caused by spinal leakage of cerebrospinal fluid (CSF) out of ventral dural tears or nerve root diverticula, or, in 2-5% of cases, through a fistula leading directly into the periradicular veins (CSF-venous fistula). In half of all patients, no CSF leak is demonstrable. A low CSF opening pressure on lumbar puncture is present in only one-third of patients; imaging studies are thus needed to confirm and localize a spinal CSF leak. Half of all patients in whom myelographic computed tomography (CT) reveals contrast medium reaching the epidural space have ventral dural tears, which tend to be located at upper thoracic spinal levels. Epidural blood patches applied under fluoroscopic or CT guidance can seal the CSF leak in 30-70% of patients, but 90% of patients with ventral dural tears will need operative closure. Some patients who have no visible epidural contrast medium on CT presumably do not have SIH, while others do, in fact, have a CSF leak from a diverticulum or a CSF-venous fistula and will need to have the site of the leak demonstrated with the aid of further studies, such as dynamic (subtraction) myelography in the lateral decubitus position. CONCLUSION: The management of patients with SIH calls for complementary imaging studies to demonstrate the causative spinal CSF leak. Often, successful treatment requires surgical closure of the leak. In view of the sparse evidence available to date, controlled studies should be performed.

Diagnostic performance and reproducibility of T2w based and diffusion weighted imaging (DWI) based PI-RADSv2 lexicon descriptors for prostate MRI.

PURPOSE: To examine the diagnostic performance of PI-RADSv2 T2w and diffusion weighted imaging (DWI) based lexicon descriptors, inter-observer agreement for descriptor assignment and diagnostic accuracy of the PI-RADSv2 assessment categories for multiparametric prostate MRI. MATERIALS AND METHODS: 176 lesions in 79 consecutive patients are analyzed, lesions are histopathologically verified by MRI-ultrasound fusion biopsy. All lesions are rated according to the PI-RADSv2 lexicon, descriptors for T2w and DWI sequences and resulting assessment categories are assigned by two independent blinded radiologists. We perform receiver-operating-characteristic analysis using the assessment categories. To analyze inter-observer agreement, we calculate weighted kappa values for assessment category assignment and unweighted kappa values for descriptor assignment. RESULTS: PI-RADSv2 assessment categories yield an area under the curve of 0.76/0.74 (radiologist 1/radiologist 2), P >0.05. Weighted kappa for agreement is 0.601 in the peripheral zone and 0.580 in the transition zone. We detect a difference in the cancer rate for PI-RADSv2 category 3 between peripheral zone (32%) and transition zone (12%), P <0.05. We obtain moderate agreement at most for descriptor assignment with kappa values ranging from 0.082 (T2w shape in the transition zone) to 0.407 (T2w signal intensity in the peripheral zone) and 0.493 (ADC pattern in the peripheral zone). Our analysis corroborates typical descriptors for benign/malignant lesions, but also reveals insights into potential pitfalls - T2w wedge shaped lesions in the peripheral zone have a considerable cancer rate, despite being labelled category 2 in the lexicon. CONCLUSION: Agreement for descriptor assignment in the PI-RADSv2 lexicon is at most moderate in our study. Typical descriptors for benign and malignant lesions are validated, whereas the discriminatory power of some descriptors is challenged. The difference in the cancer rate for PI-RADSv2 category 3 between peripheral zone and transition zone should be considered when management recommendations are linked to assessment categories in the future.

Extremity-dedicated low-field MRI shows good diagnostic accuracy and interobserver agreement for the diagnosis of the acutely injured knee.

OBJECTIVES: To estimate diagnostic accuracy and interobserver agreement of extremity-dedicated low-field magnetic resonance imaging (lfMRI) for meniscal tears, anterior cruciate ligament (ACL) tears and knee fractures. METHODS: We enrolled 62 patients with acute knee trauma and 19 patients with suspected knee fracture. Arthroscopy/Computed tomography was regarded the gold standard for cruciate ligament and meniscal tears/fractures. RESULTS: Arthroscopy showed 39 meniscal tears. Sensitivity/Specificity of lfMRI was 95.8%/97.4% (reader 1)/100%/100% (reader 2) for medial and 93.3%/100% (reader 1)/93.3%/93.6% (reader 2) for lateral meniscal tears. Sensitivity/Specificity was 100% for ACL tears and fractures. Interobserver agreement was very good. CONCLUSION: lfMRI showed reproducible high sensitivity and specificity for the diagnosis of the acutely injured knee.