Vilanilam, George K.

The translation of AI-generated brain metastases (BM) segmentation into clinical practice relies heavily on diverse, high-quality annotated medical imaging datasets. The BraTS-METS 2023 challenge has gained momentum for testing and benchmarking algorithms using rigorously annotated internationally compiled real-world datasets. This study presents the results of the segmentation challenge and characterizes the challenging cases that impacted the performance of the winning algorithms. Untreated brain metastases on standard anatomic MRI sequences (T1, T2, FLAIR, T1PG) from eight contributed international datasets were annotated in stepwise method: published UNET algorithms, student, neuroradiologist, final approver neuroradiologist. Segmentations were ranked based on lesion-wise Dice and Hausdorff distance (HD95) scores. False positives (FP) and false negatives (FN) were rigorously penalized, receiving a score of 0 for Dice and a fixed penalty of 374 for HD95. The mean scores for the teams were calculated. Eight datasets comprising 1303 studies were annotated, with 402 studies (3076 lesions) released on Synapse as publicly available datasets to challenge competitors. Additionally, 31 studies (139 lesions) were held out for validation, and 59 studies (218 lesions) were used for testing. Segmentation accuracy was measured as rank across subjects, with the winning team achieving a LesionWise mean score of 7.9. The Dice score for the winning team was 0.65 ± 0.25. Common errors among the leading teams included false negatives for small lesions and misregistration of masks in space. The Dice scores and lesion detection rates of all algorithms diminished with decreasing tumor size, particularly for tumors smaller than 100 mm3. In conclusion, algorithms for BM segmentation require further refinement to balance high sensitivity in lesion detection with the minimization of false positives and negatives. The BraTS-METS 2023 challenge successfully curated well- annotated, diverse datasets and identified common errors, facilitating the translation of BM segmentation across varied environments and providing the tools for future development of personalized volumetric reports to patients undergoing BM treatment.

There are many lesions that cause compression of nerves and vessels in the head and neck, and they can often be overlooked in the absence of adequate history or if not suspected by the radiologist. Many of these lesions require a high index of suspicion and optimal positioning for imaging. While a multimodality approach is critical in the evaluation of compressive lesions, an MRI utilizing high-resolution (heavily weighted) T2-weighted sequence is extremely useful as a starting point. In this review, we aim to discuss the radiological features of the common and uncommon compressive lesions of the head and neck which are broadly categorized into vascular, osseous, and miscellaneous etiologies.

Plasma neurofilament light is a reliable, uniform marker to estimate neuroaxonal injury and forecast all-cause mortality across major stroke types.

Introduction: White matter hyperintensities (WMH) is the most prominent neuroimaging feature of small vessel disease (SVD). SVD is thought to be a result of arteriolosclerosis, chronic hypoperfusion of white matter, selective oligodendrocyte death and degeneration of myelinated fibers. The risk of ischemic stroke is significantly increased with the presence of WMH. The presence of intracranial collaterals is a strong predictor of salvageable neurological function in acute stroke; however, the factors affecting the ability to recruit collaterals are poorly understood. We wanted to study if the presence and severity of small vessel white matter disease had an adverse effect on cerebral collateral flow in acute embolic large vessel stroke. Methods: We identified a consecutive series of adult subjects with acute embolic large vessel strokes confirmed on CT angiography (CTA) from a 10-year institutional database. Patients without a brain MRI within six months of CTA were excluded. Collateral scores of the affected hemisphere were calculated on CTA using the modified Tan scale (0-4) by two independent trained reviewers and confirmed by a neuroradiologist. The WMH was quantified using the Fazekas scale (0-3) incorporating both deep and periventricular components. Areas of restricted diffusion on brain MRI in the peri-stroke period were eliminated from the assessment of Fazekas score. Spearman’s correlation tests were performed for the association between collateral scores and WMH. Results: A total of 234 patients (mean age 70±14 years, 47% [110/234] women) satisfied the inclusion criteria. The prevalence of periventricular and deep WMH was 83% and 87% respectively. Aging had a significant impact on periventricular WMH (r=0.48, p <0.001) and deep WMH (r=0.47, p <0.001). However, collateral scores exhibited no statistically significant relationship with the presence or severity of periventricular scores (r=0.08, p=0.17) or deep WMH scores (r=0.01, p=0.77). Conclusions: Our results indicate that the presence of SVD does not affect cerebral collateral flow. The vascular niche and risk factors of SVD appear to be distinct from that of intracranial vessels responsible for collateral reperfusion of ischemic territory in acute large vessel embolic stroke.

Introduction: Poor leptomeningeal collateral flow is related to worse neurological outcome in acute ischemic stroke. The increased release of stress hormones during the first hours of acute stroke leads to inhibition of insulin secretion and increase insulin resistance, thus inducing stress hyperglycemia, which is often evident on blood glucose levels at admission. While admission hyperglycemia, independent of diabetes mellitus (DM) is associated with poor clinical outcome in ischemic stroke, studies on its effect on collateral circulation after acute cerebral embolism are largely inconclusive. Methods: We identified a consecutive series of patients aged 18 years and older from 2008 to 2018 with acute embolic large vessel stroke confirmed on CT angiography (CTA) using an institutional database. Individual CTAs were over-read, and collateral scores of the affected hemisphere were calculated using a modification of the Tan scale (0-4), by two trained readers. We reviewed the electronic medical records for presenting NIHSS, admission glucose and HbA1C levels, and prior history of DM. Spearman’s ρ, odds ratios (OR), and their 95% confidence intervals (CI) were analyzed. All tests were considered significant when P<0.05. Results: Of the 405 patients identified through the database (mean age 70 years [range, 55-85], 45% [183/405] women), 35% (142/405) had a prior history of DM. Mean admission glucose was 134±51 mg/dL, and median NIHSS at presentation was 17. Blood glucose level at admission was negatively correlated with collateral scores (ρ= -0.23, P=0.008). Specifically, glucose levels above 120 mg/dL were significantly associated (OR 1.68 CI 1.13-2.5) with poor collateral scores (≤1). However, collateral scores exhibited no statistically significant relationship with HbA1c (ρ=-0.08, P=0.15) or prior history of DM (P=0.73). Conclusions: Patients with higher admission glucose levels have poor collateral scores, while HbA1c and prior DM do not have a similar effect on collateral flow. Our results suggest that poor collateral flow may explain adverse outcomes associated with hyperglycemia in acute ischemic stroke. Determining if there is a benefit of aggressive glycemic control within the first few hours of embolic stroke warrants further study.

Introduction: Good collateral flow is associated with improved outcomes in acute embolic stroke. There is compelling data linking smoking to all types of stroke, with current smokers at a higher risk than former or never smokers. Smoking has, however, been shown to promote collaterals in coronary vascular bed in humans. It is unknown if there is a similar effect in cerebral collateral flow. We aimed to investigate if cigarette smoking was positively associated with collateral blood flow in a homogenous cohort of patients with acute large vessel occlusive embolic stroke. Methods: We identified all patients aged ≥18 with acute ischemic stroke showing embolic occlusion of ICA terminus/M1 MCA on admission CT angiography (CTA), between 2008 to 2018, through an institutional database. Patients with chronic occlusions and stenotic lesions were excluded. Individual CTAs were independently reviewed by two trained investigators to calculate collateral scores of the affected hemisphere using the modified Tan scale (0-4) as a reference. Details of cigarette smoking history were obtained from the electronic medical records. Patients were grouped into smokers (former and current smokers) and never smokers. Odds ratios (OR) and their 95% confidence intervals (CI) were estimated, and all tests were considered significant when P<0.05. Results: Of the 612 patients identified through the database, 429/612 patients (mean age 71±16 years, 45% [194/429] women) satisfied the inclusion criteria. 64.8% (278/429) had a positive history of smoking with an average of 47±12 pack-years. There was a significant association between good collateral scores (≥2) with smoking history (OR 1.74 CI 1.17-2.61, P=0.006). This trend remained significant even after comparing raw collateral scores (0-4) ( P=0.006, Spearman’s ρ=0.22). However, number of pack-years and collateral scores exhibited no statistically significant relationship (P=0.6). Conclusions: Our study shows that patients who have ever smoked have higher intracranial collateral vessel scores in the event of an acute embolic stroke. Increased impedance and reduced cerebral blood flow resulting from cigarette smoking may stimulate collateral recruitment in cerebral pial circulation, similar to the effect in coronary bed.