Senior Neuroradiologist Research Articles

Deep Learning-Based ASPECTS Algorithm Enhances Reader Performance and Reduces Interpretation Time.

ASPECTS is a long-standing and well documented selection criteria for acute ischemic stroke treatment, however, the interpretation of ASPECTS is a challenging and time-consuming task for physicians with significant interobserver variabilities. We conducted a multi-reader, multi-case study in which readers assessed ASPECTS without and with the support of a deep learning (DL)-based algorithm in order to analyze the impact of the software on clinicians' performance and interpretation time. A total of 200 NCCT scans from 5 clinical sites (27 scanner models, 4 different vendors) were retrospectively collected. Reference standard was established through the consensus of three expert neuroradiologists who had access to baseline CTA and CTP data. Subsequently, eight additional clinicians (four typical ASPECTS reader and four senior neuroradiologists) analyzed the NCCT scans without and with the assistance of CINA-ASPECTS (Avicenna.AI, La Ciotat, France), a DLbased FDA-cleared and CE-marked algorithm designed to automatically compute ASPECTS. Differences were evaluated in both performance and interpretation time between the assisted and unassisted assessments. With software aid, readers demonstrated increased region-based accuracy from 72.4% to 76.5% (p<0.05), and increased ROC AUC from 0.749 to 0.788 (p<0.05). Notably, all readers exhibited an improved ROC AUC when utilizing the software. Moreover, use of the algorithm improved the score-based inter-observer reliability and correlation coefficient of ASPECTS evaluation by 0.222 and 0.087 (p<0.0001), respectively. Additionally, the readers' mean time spent analyzing a case was significantly reduced by 6% (p<0.05) when aided by the algorithm. With the assistance of the algorithm, readers' analyses were not only more accurate but also faster. Additionally, the overall ASPECTS evaluation exhibited greater consistency, less variabilities and higher precision compared to the reference standard. This novel tool has the potential to enhance patient selection for appropriate treatment by enabling physicians to deliver accurate and timely diagnosis of acute ischemic stroke. ASPECTS = Alberta Stroke Program Early Computed Tomography Score; DL = Deep Learning; EIC = Early Ischemic Changes; ICC = Intraclass Correlation Coefficient; IS = Ischemic Stroke; ROC AUC = Receiver Operating Characteristics Area Under the Curve.

Precise Image-level Localization of Intracranial Hemorrhage on Head CT Scans with Deep Learning Models Trained on Study-level Labels.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To develop a highly generalizable weakly supervised model to automatically detect and localize image- level intracranial hemorrhage (ICH) using study-level labels. Materials and Methods In this retrospective study, the proposed model was pretrained on the image-level RSNA dataset and fine-tuned on a local dataset using attention-based bidirectional long-short-term memory networks. This local training dataset included 10,699 noncontrast head CT scans from 7469 patients with ICH study-level labels extracted from radiology reports. Model performance was compared with that of two senior neuroradiologists on 100 random test scans using the McNemar test, and its generalizability was evaluated on an external independent dataset. Results The model achieved a positive predictive value (PPV) of 85.7% (95% CI: [84.0%, 87.4%]) and an AUC of 0.96 (95% CI: [0.96, 0.97]) on the held-out local test set (n = 7243, 3721 female) and 89.3% (95% CI: [87.8%, 90.7%]) and 0.96 (95% CI: [0.96, 0.97]), respectively, on the external test set (n = 491, 178 female). For 100 randomly selected samples, the model achieved performance on par with two neuroradiologists, but with a significantly faster (P < .05) diagnostic time of 5.04 seconds per scan (versus 86 seconds and 22.2 seconds for the two neuroradiologists, respectively). The model's attention weights and heatmaps visually aligned with neuroradiologists' interpretations. Conclusion The proposed model demonstrated high generalizability and high PPVs, offering a valuable tool for expedited ICH detection and prioritization while reducing false-positive interruptions in radiologists' workflows. ©RSNA, 2024.

Diagnostic accuracy and radiological validation of intracerebral hemorrhage diagnosis in the Swedish Stroke Register (Riksstroke).

National quality registries for stroke care operate under the assumption that the included patients are correctly diagnosed. We aimed to validate the clinical diagnosis of spontaneous intracerebral hemorrhage (ICH) in Riksstroke (RS) by evaluating radiological data from a large, unselected ICH population. We conducted a retrospective, multicenter study including all ICH patients registered in RS between 2016 and 2020 residing in Skåne County in Sweden (1.41 million inhabitants). Radiological data from first imaging were evaluated for the presence of spontaneous ICH. Other types of bleeds were registered if a spontaneous ICH was not identified on imaging. The radiological evaluation was independently performed by one radiology fellow and one senior neuroradiologist. Between 2016 and 2020, 1784 ICH cases were registered in RS, of which 1655 (92.8%) had a radiological diagnosis consistent with spontaneous ICH. In the 129 (7.2%) remaining cases, the radiological diagnosis was instead traumatic bleed (n = 80), subarachnoid hemorrhage (n = 15), brain tumor bleed (n = 14), ischemic lesion with hemorrhagic transformation (n = 14), ischemic lesion (n = 3), or no bleed at all (n = 3). There was a higher degree of incorrect coding in the older age groups. At radiological evaluation, 92.8% of ICH diagnoses in RS were consistent with spontaneous ICH, yielding a high rate of agreement that strengthens the validity of the diagnostic accuracy in the register, justifying the use of high coverage quality register data for epidemiological purposes. The most common coding error was traumatic bleeds that were classified as spontaneous ICH.

Differentiating Low-Grade from High-Grade Intracranial Ependymomas: Comparison of Dynamic Contrast-Enhanced MRI and Diffusion-Weighted Imaging.

The aim of this study was to determine the diagnostic value of fractional plasma volume derived from dynamic contrast-enhanced perfusion MR imaging versus ADC, obtained from DWI in differentiating between grade 2 (low-grade) and grade 3 (high-grade) intracranial ependymomas. A hospital database was created for the period from January 2013 through June 2022, including patients with histologically-proved ependymoma diagnosis with available dynamic contrast-enhanced MR imaging. Both dynamic contrast-enhanced perfusion and DWI were performed on each patient using 1.5T and 3T scanners. Fractional plasma volume maps and ADC maps were calculated. ROIs were defined by a senior neuroradiologist manually by including the enhancing tumor on every section and conforming a VOI to obtain the maximum value of fractional plasma volume (Vpmax) and the minimum value of ADC (ADCmin). A Mann-Whitney U test at a significance level of corrected P = .01 was used to evaluate the differences. Additionally, receiver operating characteristic curve analysis was applied to assess the sensitivity and specificity of Vpmax and ADCmin values. A total of 20 patients with ependymomas (10 grade 2 tumors and 10 grade 3 tumors) were included. Vpmax values for grade 3 ependymomas were significantly higher (P < .002) than those for grade 2. ADCmin values were overall lower in high-grade lesions. However, no statistically significant differences were found (P = .12114). As a dynamic contrast-enhanced perfusion MR imaging metric, fractional plasma volume can be used as an indicator to differentiate grade 2 and grade 3 ependymomas. Dynamic contrast-enhanced perfusion MR imaging plays an important role with high diagnostic value in differentiating low- and high-grade ependymoma.

Incidence and predictors of intracranial hemorrhage after intravenous thrombolysis with tenecteplase.

Despite its increasing use, there are limited data on the risk of intracranial hemorrhage (ICH) after intravenous thrombolysis with tenecteplase in the setting of acute ischemic stroke. Our aim was to investigate the incidence and predictors of ICH after tenecteplase administration. We reviewed data from the prospective ongoing multicenter TETRIS (Tenecteplase Treatment in Ischemic Stroke) registry. Patients with available day-1 imaging were included in this study. Clinical, imaging and biological variables were collected. Follow-up imaging performed 24 h after IVT was locally reviewed by senior neuroradiologists and neurologists. The incidence of parenchymal hematoma (PH) and any ICH were investigated. Potential predictors of PH and any ICH were assessed in multivariable logistic regressions. Subgroup analyses focusing on patients intended for endovascular treatment were performed. PH and any ICH occurred in 126/1321 (incidence rate: 9.5%, 95% CI 8.1-11.2) and 521/1321 (39.4%, 95% CI 36.8-42.1) patients, respectively. Symptomatic ICH was observed in 77/1321 (5.8%; 95% CI 4.7-7.2). PH occurrence was significantly associated with poorer functional outcomes (p < 0.0001) and death (p < 0.0001) after 3 months. Older age (aOR = 1.03; 95% CI 1.01-1.05), male gender (aOR = 2.07; 95% CI 1.28-3.36), a history of hypertension (aOR = 2.08; 95% CI 1.19-3.62), a higher baseline NIHSS (aOR = 1.07; 95% CI 1.03-1.10) and higher admission blood glucose level (aOR = 1.12; 95% CI 1.05-1.19) were independently associated with PH occurrence. Similar associations were observed in the subgroup of patients intended for endovascular treatment. We quantified the incidence of ICH after IVT with tenecteplase in a real-life prospective registry and determined independent predictors of ICH. These findings allow to identify patients at high risk of ICH.

Unravelling pituitary tumours in medically treated patients with acromegaly: the impact of systematic MRI reassessment.

In acromegaly, the primary tumor is usually found during magnetic resonance imaging (MRI) of the pituitary gland. A remnant tumor after surgery is, however, harder to depict. When a tumor is missed, the remaining option is usually lifelong pharmacological treatment. To identify tumors by reassessment of all available MRI scans in pharmacologically treated patients, operated or not, and to compare our results with the routine MRI reports. Adult patients diagnosed with acromegaly and managed at a tertiary care center between 2005 and 2021 and currently on pharmacological treatment were included. MRI scans were evaluated in a standardized manner and classified independently by a radiologist and an endocrinologist into "certain," "suspected," or "no tumor." In case of disagreement, consensus was achieved with a senior neuroradiologist. The results were compared using the clinical radiologists' routine MRI reports. We identified certain and suspected tumors in 29/74 and 36/74 patients, respectively. No tumor was identified in nine patients. In five of these, no MRI contrast agent was given. Discrepancy between our results and the routine MRI reports was found in 31/74 patients (P = 0.01). In 22 patients, the routine reports described no tumor while we identified certain tumors in 2/22 patients and suspected tumors in 13/22 patients. In most patients with pharmacologically treated acromegaly, we identified a certain or suspected pituitary tumor. These findings were more frequent compared to the routine MRI reports. Based on our results, patients will be considered for a change in long-term treatment modality.

Assessment of fetal corpus callosum biometry by 3D super-resolution reconstructed T2-weighted magnetic resonance imaging.

To assess the accuracy of corpus callosum (CC) biometry, including sub-segments, using 3D super-resolution fetal brain MRI (SR) compared to 2D or 3D ultrasound (US) and clinical low-resolution T2-weighted MRI (T2WS). Fetal brain biometry was conducted by two observers on 57 subjects [21-35 weeks of gestational age (GA)], including 11 cases of partial CC agenesis. Measures were performed by a junior observer (obs1) on US, T2WS and SR and by a senior neuroradiologist (obs2) on T2WS and SR. CC biometric regression with GA was established. Statistical analysis assessed agreement within and between modalities and observers. This study shows robust SR to US concordance across gestation, surpassing T2WS. In obs1, SR aligns with US, except for genu and CC length (CCL), enhancing splenium visibility. In obs2, SR closely corresponds to US, differing in rostrum and CCL. The anterior CC (rostrum and genu) exhibits higher variability. SR's regression aligns better with literature (US) for CCL, splenium and body than T2WS. SR is the method with the least missing values. SR yields CC biometry akin to US (excluding anterior CC). Thanks to superior 3D visualization and better through plane spatial resolution, SR allows to perform CC biometry more frequently than T2WS.

Preliminary data on artificial intelligence tool in magnetic resonance imaging assessment of degenerative pathologies of lumbar spine.

To evaluate the ability of an artificial intelligence (AI) tool in magnetic resonance imaging (MRI) assessment of degenerative pathologies of lumbar spine using radiologist evaluation as a gold standard. Patients with degenerative pathologies of lumbar spine, evaluated with MRI study, were enrolled in a retrospective study approved by local ethical committee. A comprehensive software solution (CoLumbo; SmartSoft Ltd., Varna, Bulgaria) designed to label the segments of the lumbar spine and to detect a broad spectrum of degenerative pathologies based on a convolutional neural network (CNN) was employed, utilizing an automatic segmentation. The AI tool efficacy was compared to data obtained by a senior neuroradiologist that employed a semiquantitative score. Chi-square test was used to assess the differences among groups, and Spearman's rank correlation coefficient was calculated between the grading assigned by radiologist and the grading obtained by software. Moreover, agreement was assessed between the value assigned by radiologist and software. Ninety patients (58 men; 32 women) affected with degenerative pathologies of lumbar spine and aged from 60 to 81years (mean 66years) were analyzed. Significant correlations were observed between grading assigned by radiologist and the grading obtained by software for each localization. However, only when the localization was L2-L3, there was a good correlation with a coefficient value of 0.72. The best agreements were obtained in case of L1-L2 and L2-L3 localizations and were, respectively, of 81.1% and 72.2%. The lowest agreement of 51.1% was detected in case of L4-L5 locations. With regard canal stenosis and compression, the highest agreement was obtained for identification of in L5-S1 localization. AI solution represents an efficacy and useful toll degenerative pathologies of lumbar spine to improve radiologist workflow.

CTA Supplemented by CTP Increases Interrater Reliability and Endovascular Treatment Use in Patients with Aneurysmal SAH.

Cerebral vasospasm is a common complication of aneurysmal SAH and remains a risk factor for delayed cerebral ischemia and poor outcome. The interrater reliability of CTA in combination with CTP has not been sufficiently studied. We aimed to investigate the reliability of CTA alone and in combination with CTP in the detection of cerebral vasospasm and the decision to initiate endovascular treatment. This is a retrospective single-center study including patients treated for aneurysmal SAH. Inclusion criteria were a baseline CTA and follow-up imaging including CTP due to suspected vasospasm. Three neuroradiologists were asked to grade 15 intracranial arterial segments in 71 cases using a tripartite scale (no, mild <50%, or severe >50% vasospasm). Raters further evaluated whether endovascular treatment should be indicated. The ratings were performed in 2 stages with a minimum interval of 6 weeks. The first rating included only CTA images, whereas the second rating additionally encompassed CTP images. All raters were blinded to any clinical information of the patients. Interrater reliability for per-segment analysis of vessels was highly variable (κ = 0.16-0.61). We observed a tendency toward higher interrater reliability in proximal vessel segments, except for the ICA. CTP did not improve the reliability for the per-segment analysis. When focusing on senior raters, the addition of CTP images resulted in higher interrater reliability for severe vasospasm (κ = 0.28; 95% CI, 0.10-0.46 versus κ = 0.46; 95% CI, 0.26-0.66) and subsequently higher concordance (κ = 0.23; 95% CI, -0.01-0.46 versus κ = 0.73; 95% CI, 0.55-0.91) for the decision of whether endovascular treatment was indicated. CTA alone offers only low interrater reliability in the graduation of cerebral vasospasm. However, using CTA in combination with CTP might help, especially senior neuroradiologists, to increase the interrater reliability to identify severe vasospasm following aneurysmal SAH and to increase the reliability regarding endovascular treatment decisions.

Classification of traumatic injury to the dural venous sinus using CT venography.

Cerebral venous sinus thrombosis (CVST) is an underrecognized cause of morbidity in acute traumatic brain injury (TBI). Radiologic diagnosis is challenging in the setting of concurrent extra-axial injury and a lack of standardized diagnostic criteria. The prevalence of traumatic thrombosis versus compression is unknown. Treatment with anticoagulation is often determined by the appropriate classification of the type of traumatic venous injury. We developed a two-part radiologic grading method for standardized assessment of traumatic CVST based on (1) the degree of flow limitation through the affected sinus and (2) the location of venous pathology (ie, external compression vs. intrinsic thrombosis) based on computed tomography venography. We applied this grading method to a retrospective cohort of TBI patients presenting to a Level 1 Trauma center. Chart review was performed to identify potential clinical correlates. A senior neuroradiologist graded the entire cohort and a random subsample was selected for blinded rating by two independent neuroradiologists. Seventy-six of 221 patients were identified for inclusion after excluding nontraumatic mechanisms. Seven unique grades were employed to characterize the full extent of venous injuries. The plurality of patients from the cohort (43/76 = 43.4%) suffered compressive injuries. Inter-rater reliability was moderate for the combined grade, kappa = 0.48, p<.05, and substantial for the flow limitation component, kappa = 0.69, p<.05. We introduce a standardized two-part classification system for traumatic venous sinus injury with moderate-substantial inter-rater reliability. Compressive injuries were more common than thrombotic injuries. Further prospective work is needed to validate the clinical significance of this classification system.

Accuracy of Noncontrast T2 SPACE in Active MS Cord Lesion Detection.

The diagnosis of active MS lesions is often based on postgadolinium T1-weighted MR imaging. Recent studies suggest a risk of IV gadolinium to patients, predominantly based on gadolinium deposition in tissue. Noncontrast sequences have shown promise in MS diagnosis, but none differentiate acute from chronic MS lesions. We hypothesized that 3D T2 sampling perfection with application-optimized contrasts by using different flip angle evolution (SPACE) MR imaging can help detect and differentiate active-versus-chronic MS lesions without the need for IV contrast. In this single-center retrospective study, 340 spinal MR imaging cases of MS were collected in a 24-month period. Two senior neuroradiologists blindly and independently reviewed postcontrast T1-weighted sagittal and T2-SPACE sagittal images for the presence of MS lesions, associated cord expansion/atrophy on T2-SPACE, and enhancement on postcontrast T1WI. Discrepancies were resolved by consensus between the readers. Sensitivity, specificity, and accuracy of T2-SPACE compared with postcontrast T1WI were computed, and interobserver agreement was calculated. The sensitivity of lesion detection on T2-SPACE was 85.71%, 95% CI, 63.66%-96.95%; with a specificity of 93.52%, 95% CI, 90.06%-96.05%; and an accuracy of 92.99%, 95% CI, 89.58%-95.56. Additionally, 16/21 (84.2%) acute enhancing cord lesions showed cord expansion on T2-SPACE. The interobserver agreement was 92%. Our study shows that T2-SPACE facilitates noncontrast detection of acute MS lesions with high accuracy compared with postcontrast T1WI and with high interobserver agreement. The lack of gadolinium use provides an advantage, bypassing any potential adverse effects of repetitive contrast administration.

High diagnostic accuracy of T2FLAIR at 3 T in the detection of optic nerve head edema in acute optic neuritis.

Optic nerve head edema (ONHE) detected by fundoscopy is observed in one-third of patients presenting optic neuritis (ON). While ONHE is an important semiological feature, the correlation between ONHE and optic nerve head MRI abnormalities (ONHMA), sometimes called "optic nerve head swelling," remains unknown. Our study aimed to assess the diagnostic accuracy of T2 fluid-attenuated inversion recovery (FLAIR) MRI sequence in detecting ONHE in patients with acute ON. In the present single-center study, data were extracted from two prospective cohort studies that consecutively included adults with a first episode of acute ON treated between 2015 and 2020. Two experienced readers blinded to study data independently analyzed imaging. A senior neuroradiologist resolved any discrepancies. The primary judgment criterion of ONHMA was assessed as optic nerve head high signal intensity on gadolinium-enhanced T2FLAIR MRI sequence. Its diagnostic accuracy was evaluated with both the gold standard of ONHE on fundus photography (FP) and peripapillary retinal nerve fiber layer thickening on optic coherence tomography (OCT). A total of 102 patients were included, providing 110 affected and 94 unaffected optic nerves. Agreement was high between the different modalities: 92% between MRI and FP (k= 0.77, 95% CI: 0.67-0.88) and 93% between MRI and OCT (k= 0.77, 95% CI: 0.67-0.87). MRI sensitivity was 0.84 (95% CI: 0.70-0.93) and specificity was 0.94 (95% CI: 0.89-0.97) when compared with the FP. Optic nerve head high T2FLAIR signal intensity corresponds indeed to the optic nerve head edema diagnosed by the ophthalmologists. MRI is a sensitive tool for detecting ONHE in patients presenting acute ON. In patients with optic neuritis the high T2FLAIR (fluid-attenuated inversion recovery) signal intensity of the optic nerve head corresponds indeed to optic nerve head edema, which is a useful feature in optic neuritis etiological evaluation and treatment. Optic nerve head edema is a prominent clinical feature of acute optic neuritis and is usually diagnosed during dilated or non-dilated eye fundus examination. Agreement was high between magnetic resonance imaging, fundus photography, and optical coherence tomography. Optic nerve head high T2 fluid attenuation inversion recovery signal intensity is a promising detection tool for optic nerve head edema in patients presenting acute optic neuritis.

Integrating navigation assistance for redirecting freehanded spinal instrumentation: experience and technique

Retrospective review of all spinal fusions > 3 levels involving the thoracolumbar and/or sacroiliac at a single institution, by a single surgeon between 3/12/2020 and 8/13/2021 were reviewed. All screws that were secondarily navigated after identified as misdirected on intraoperative CT scan were included. Neuromonitoring reports were culled for mA threshold to triggered EMG response for all redirected screws. Intraoperative, post-de novo screw placement images (fluoroscopy scout and intraoperative CT) and post-redirection intraoperative scoliosis films and post-operative scoliosis films were independently reviewed by a senior neuroradiologist. Fifty redirected screws in the thoracic, lumbar, sacral, and ilium were identified as misdirected and redirected via navigation. The new trajectory of all screws was confirmed satisfactory by independent review between a senior neuroradiologist and neurosurgeon. Four screws could not be verified by post-operative imaging (4/50, 8%). All triggered EMG stimulated > 15 mA. No screws required return to the operating room for revision. No patients experienced a post-operative deficit. Redirection of misdirected thoracolumbar and sacroiliac screws can be performed using intraoperative CT and navigation as a means to detect and directly visualize appropriate placement.

Impact of rater experience and referral question on detecting magnetic resonance imaging features of idiopathic intracranial hypertension.

In idiopathic intracranial hypertension (IIH), magnetic resonance imaging (MRI) features are promising diagnostic markers, but the impact of rater experience and the specific referral question is unknown. From the Vienna Idiopathic Intracranial Hypertension database, patients were included with definitive IIH and routine cranial MRI performed during diagnostic work-up. Frequencies of partial empty sella (ES), optic nerve sheath distension (ONSD), optic nerve tortuosity (ONT), posterior globe flattening (PGF) and transverse sinus stenosis (TSS) were compared in three settings: (i) real-world rating, (ii) junior neuroradiologist without special IIH training and (iii) senior neuroradiologist with experience in IIH imaging (gold standard). Magnetic resonance imaging scans of 84 IIH patients (88% female, mean age 33.5 years) were evaluated. By gold standard, ONSD was the most frequent (64.3%) followed by TSS (60.0%), ONT (46.4%), ES (44.4%) and PGF (23.8%). Compared to the gold standard, IIH features were described significantly less frequently in routine MRI reports (ONSD 28.6%, ONT 13.1%, PGF 4.8%, TSS 42.9%, p < 0.01 respectively) except for ES (42.9%, p = 0.9). A specific referral question regarding IIH increased detection rates in routine reports, but rates remained significantly lower than by gold standard. In contrast, a rating by a neuroradiologist without special training produced significantly higher frequencies of ONSD (81.0%, p < 0.01) and ONT (60.7%, p < 0.01) but not of ES (47.6%), PGF (29.8%) and TSS (68.1%). Idiopathic intracranial hypertension MRI features are underestimated in routine MRI reports and partly overcalled by less experienced neuroradiologists, driven by features less well known or methodologically difficult. Reevaluation of MRI scans by an experienced rater (and to a lesser degree a specific referral question) improves diagnostic accuracy.

Opportunistic CT Screening-Machine Learning Algorithm Identifies Majority of Vertebral Compression Fractures: A Cohort Study.

Vertebral compression fractures (VCF) are common in patients older than 50 years but are often undiagnosed. Zebra Medical Imaging developed a VCF detection algorithm, with machine learning, to detect VCFs from CT images of the chest and/or abdomen/pelvis. In this study, we evaluated the diagnostic performance of the algorithm in identifying VCF. We conducted a blinded validation study to estimate the operating characteristics of the algorithm in identifying VCFs using previously completed CT scans from 1200 women and men aged 50 years and older at a tertiary-care center. Each scan was independently evaluated by two of three neuroradiologists to identify and grade VCF. Disagreements were resolved by a senior neuroradiologist. The algorithm evaluated the CT scans in a separate workstream. The VCF algorithm was not able to evaluate CT scans for 113 participants. Of the remaining 1087 study participants, 588 (54%) were women. Median age was 73 years (range 51-102 years; interquartile range 66-81). For the 1087 algorithm-evaluated participants, the sensitivity and specificity of the VCF algorithm in diagnosing any VCF were 0.66 (95% confidence interval [CI] 0.59-0.72) and 0.90 (95% CI 0.88-0.92), respectively, and for diagnosing moderate/severe VCF were 0.78 (95% CI 0.70-0.85) and 0.87 (95% CI 0.85-0.89), respectively. Implementing this VCF algorithm within radiology systems may help to identify patients at increased fracture risk and could support the diagnosis of osteoporosis and facilitate appropriate therapy. © 2023 Amgen, Inc. JBMR Plus published by Wiley Periodicals LLC on behalf of American Society for Bone and Mineral Research.

Sex-Specific Patterns of Cerebral Atrophy and Enlarged Perivascular Spaces in Patients with Cerebral Amyloid Angiopathy and Dementia.

Cerebral amyloid angiopathy is characterized by amyloid β deposition in leptomeningeal and superficial cortical vessels. Cognitive impairment is common and may occur independent of concomitant Alzheimer disease neuropathology. It is still unknown which neuroimaging findings are associated with dementia in cerebral amyloid angiopathy and whether they are modulated by sex. This study compared MR imaging markers in patients with cerebral amyloid angiopathy with dementia or mild cognitive impairment or who are cognitively unimpaired and explored sex-specific differences. We studied 58 patients with cerebral amyloid angiopathy selected from the cerebrovascular and memory outpatient clinics. Clinical characteristics were collected from clinical records. Cerebral amyloid angiopathy was diagnosed on MR imaging on the basis of the Boston criteria. Visual rating scores for atrophy and other imaging features were independently assessed by 2 senior neuroradiologists. Medial temporal lobe atrophy was higher for those with cerebral amyloid angiopathy with dementia versus those cognitively unimpaired (P = .015), but not for those with mild cognitive impairment. This effect was mainly driven by higher atrophy in men with dementia, compared with women with and without dementia (P = .034, P = .012; respectively) and with men without dementia (P = .012). Enlarged perivascular spaces in the centrum semiovale were more frequent in women with dementia versus men with and without dementia (P = .021, P = .011; respectively) and women without dementia (P = .011). Medial temporal lobe atrophy was more prominent in men with dementia, whereas women showed a higher number of enlarged perivascular spaces in the centrum semiovale. Overall, this finding suggests differential pathophysiologic mechanisms with sex-specific neuroimaging patterns in cerebral amyloid angiopathy.

The value of convolutional neural networks-based deep learning model in differential diagnosis of space-occupying brain diseases.

It is still a challenge to differentiate space-occupying brain lesions such as tumefactive demyelinating lesions (TDLs), tumefactive primary angiitis of the central nervous system (TPACNS), primary central nervous system lymphoma (PCNSL), and brain gliomas. Convolutional neural networks (CNNs) have been used to analyze complex medical data and have proven transformative for image-based applications. It can quickly acquire diseases' radiographic features and correct doctors' diagnostic bias to improve diagnostic efficiency and accuracy. The study aimed to assess the value of CNN-based deep learning model in the differential diagnosis of space-occupying brain diseases on MRI. We retrospectively analyzed clinical and MRI data from 480 patients with TDLs (n = 116), TPACNS (n = 64), PCNSL (n = 150), and brain gliomas (n = 150). The patients were randomly assigned to training (n = 240), testing (n = 73), calibration (n = 96), and validation (n = 71) groups. And a CNN-implemented deep learning model guided by clinical experts was developed to identify the contrast-enhanced T1-weighted sequence lesions of these four diseases. We utilized accuracy, sensitivity, specificity, and area under the curve (AUC) to evaluate the performance of the CNN model. The model's performance was then compared to the neuroradiologists' diagnosis. The CNN model had a total accuracy of 87% which was higher than senior neuroradiologists (74%), and the AUC of TDLs, PCNSL, TPACNS and gliomas were 0.92, 0.92, 0.89 and 0.88, respectively. The CNN model can accurately identify specific radiographic features of TDLs, TPACNS, PCNSL, and gliomas. It has the potential to be an effective auxiliary diagnostic tool in the clinic, assisting inexperienced clinicians in reducing diagnostic bias and improving diagnostic efficiency.

Automated CT angiography collateral scoring in anterior large vessel occlusion stroke: A multireader study.

Computed tomography (CT) angiography collateral score (CTA-CS) is an important clinical outcome predictor following mechanical thrombectomy for ischemic stroke with large vessel occlusion (LVO). The present multireader study aimed to evaluate the performance of e-CTA software for automated assistance in CTA-CS scoring. Brain CTA images of 56 patients with anterior LVO were retrospectively processed. Twelve readers of various clinical training, including junior neuroradiologists, senior neuroradiologists, and neurologists graded collateral flow using visual CTA-CS scale in two sessions separated by a washout period. Reference standard was the consensus of three expert readers. Duration of reading time, inter-rater reliability, and statistical comparison of readers' performance metrics were analyzed between the e-CTA assisted and unassisted sessions. e-CTA assistance resulted in significant increase in mean accuracy (58.6% to 67.5%, p = 0.003), mean F1 score (0.574 to 0.676, p = 0.002), mean precision (58.8% to 68%, p = 0.007), and mean recall (58.7% to 69.9%, p = 0.002), especially with slight filling deficit (CTA-CS 2 and 3). Mean reading time was reduced across all readers (103.4 to 59.7 s, p = 0.001), and inter-rater agreement in CTA-CS assessment was increased (Krippendorff's alpha 0.366 to 0.676). Optimized occlusion laterality detection was also noted with mean accuracy (92.9% to 96.8%, p = 0.009). Automated assistance for CTA-CS using e-CTA software provided helpful decision support for readers in terms of improving scoring accuracy and reading efficiency for physicians with a range of experience and training backgrounds and leading to significant improvements in inter-rater agreement.

An MRI-based joint model of radiomics and spatial distribution differentiates autoimmune encephalitis from low-grade diffuse astrocytoma

The differential diagnosis between autoimmune encephalitis and low-grade diffuse astrocytoma remains challenging. We aim to develop a quantitative model integrating radiomics and spatial distribution features derived from MRI for discriminating these two conditions. In our study, we included 188 patients with confirmed autoimmune encephalitis (n = 81) and WHO grade II diffuse astrocytoma (n = 107). Patients with autoimmune encephalitis (AE, n = 59) and WHO grade II diffuse astrocytoma (AS, n = 79) were divided into training and test sets, using stratified sampling according to MRI scanners. We further included an independent validation set (22 patients with AE and 28 patients with AS). Hyperintensity fluid-attenuated inversion recovery (FLAIR) lesions were segmented for each subject. Ten radiomics and eight spatial distribution features were selected via the least absolute shrinkage and selection operator (LASSO), and joint models were constructed by logistic regression for disease classification. Model performance was measured in the test set using the area under the receiver operating characteristic (ROC) curve (AUC). The discrimination performance of the joint model was compared with neuroradiologists. The joint model achieved better performance (AUC 0.957/0.908, accuracy 0.914/0.840 for test and independent validation sets, respectively) than the radiomics and spatial distribution models. The joint model achieved lower performance than a senior neuroradiologist (AUC 0.917/0.875) but higher performance than a junior neuroradiologist (AUC 0.692/0.745) in the test and independent validation sets. The joint model of radiomics and spatial distribution from a single FLAIR could effectively classify AE and AS, providing clinical decision support for the differential diagnosis between the two conditions.

Searching for underlying atrial fibrillation using artificial intelligence-assisted MRI images from ischemic stroke patients

Abstract Background Nearly 30% of ischemic strokes are attributed to atrial fibrillation (AF) related cardiac thromboembolism. This subset of strokes are associated with higher recurrence rate, morbidity and mortality. As a result, current recommendation suggests a cardiac monitoring of variable intensity and duration to search for unknown AF after a stroke so that optimal secondary prevention could be administered. Unfortunately, it is reported that 24h Holter ECG monitoring for underlying AF is completed routinely in only a minority of stroke centers, making a huge gap between AF detection rate in clinical practice and estimated actual data (2–3% vs 23–30%). So, it is obvious that paroxysmal AF remains undetected in a significant proportion of patients after a stroke and an early and effective identification approach of cardiac disturbances is needed. Purpose We aim to develop an innovative end-to-end artificial intelligence model to rapidly search for atrial fibrillation in post ischemic stroke patients using MRI imaging data. Methods 489 patients diagnosed with ischemic stroke, 174 of which had a prior known history of AF or new AF detected by intensive ECG monitoring poststroke, from January 2018 to October 2021 were enrolled in this study. The MRI images obtained after their admission in the stroke unit were meticulously evaluated and acute ischemic stroke lesions on DWI sequences were segmented and confirmed by two senior neuroradiologists independently. These processed images were randomly split into training (n=315) and testing set (n=174) for the implementation and validation of our AI screening model. And the algorithm was based on the combination of radiomic features and semantic features extracted from convolutional neural network (CNN). The model performance was evaluated by accuracy, recall, precision, F-Measure and AUC. Also, heatmaps which indicate the attention mechanism of the model were generated for interpreting underlying patterns. Results The AUC of our model (COM) reached 0.8 (Figure 1A). Also, the algorithm yielded values for the accuracy, recall, precision and F-Measure of 70%, 92.5%, 63.8% and 75.5%, respectively, which showed satisfactory classification results. As to the interpretability of the AI model, we found that more “attention” was paid to the main lesions (Figure 1B) and radiomic features which indicated the minimum gray level intensity and the sphericity of the lesions (Figure 1C, D) were crucial to the classifier. Conclusions Our work revealed a potential correlation between brain ischemic lesion pattern on DWI images and underlying etiology of AF. Moreover, the AI model we developed may serve as a rapid screening tool for AF in clinical practice of stroke units. Funding Acknowledgement Type of funding sources: Public grant(s) – National budget only. Main funding source(s): National Natural Science Foundation of China