Clinical and radiologic characteristics associated with multiple sclerosis misdiagnosis at a tertiary referral center in the United States.

BACKGROUND: Misdiagnosis of multiple sclerosis (MS) is common and can have harmful effects on patients and healthcare systems. Identification of factors associated with misdiagnosis may aid development of prevention strategies. OBJECTIVE: To identify clinical and radiological predictors of MS misdiagnosis. METHODS: We retrospectively reviewed medical records of all patients who were referred to Johns Hopkins MS Center from January 2018 to June 2019. Patients who carried a diagnosis of MS were classified as correctly diagnosed or misdiagnosed with MS by the Johns Hopkins clinician. Demographics, clinical, laboratory, and radiologic data were collected. Differences between the two groups were evaluated, and a regression model was constructed to identify predictors of misdiagnosis. RESULTS: Out of 338 patients who were previously diagnosed with MS, 41 (12%) had been misdiagnosed. An alternative diagnosis was confirmed in 28 (68%) of the misdiagnosed patients; cerebrovascular disease was the most common alternate diagnosis. Characteristics associated with misdiagnosis were female sex (odds ratio (OR): 5.81 (95% confidence interval (CI): 1.60, 21.05)) and non-specific brain magnetic resonance imaging (MRI) lesions (OR: 7.66 (3.42, 17.16)). CONCLUSION: Misdiagnosis is a frequent problem in MS care. Non-specific brain lesions were the most significant predictor of misdiagnosis. Interventions aimed to reduce over-reliance on imaging findings and misapplication of the McDonald criteria may prevent MS misdiagnosis.

Foundations of Lesion Detection Using Machine Learning in Clinical Neuroimaging.

This chapter describes technical considerations and current and future clinical applications of lesion detection using machine learning in the clinical setting. Lesion detection is central to neuroradiology and precedes all further processes which include but are not limited to lesion characterization, quantification, longitudinal disease assessment, prognosis, and prediction of treatment response. A number of machine learning algorithms focusing on lesion detection have been developed or are currently under development which may either support or extend the imaging process. Examples include machine learning applications in stroke, aneurysms, multiple sclerosis, neuro-oncology, neurodegeneration, and epilepsy.

Relevance of antinuclear antibody in diagnosis and characteristics of multiple sclerosis.

BACKGROUND: Criteria for multiple sclerosis (MS) diagnosis rely upon clinical and paraclinical data that are supportive of MS in the absence of a better explanation. Patients referred for consideration of a MS diagnosis often undergo an extensive serologic workup including antinuclear antibody (ANA) testing, even when an individual already meets diagnostic criteria for MS. It is unclear whether ANA serostatus is associated with clinical outcomes in MS. The present study aims to determine if ANA seropositivity in those referred with concern for MS differs in those who meet 2017 revised McDonald criteria compared to those who did not receive a diagnosis of MS. Associations between ANA seropositivity and clinical or radiological phenotype of MS patients are also explored. METHODS: The cohort included people at least 18 years old, referred to our tertiary care MS center with concern for MS (regardless of prior diagnosis) who had an ANA test with known titer completed within one year of first evaluation. Electronic health record (EHR) charts were manually reviewed, and MRIs underwent blinded review by a radiologist with training in neuroradiology. Diagnosis of MS was determined by a neuroimmunologist and was based on 2017 revised McDonald Criteria. Results are reported as odds ratios from multivariable logistic regression analyses adjusted for age, sex at birth, race, smoking history, personal history of comorbid autoimmune conditions, and family history of autoimmunity. Within the MS cohort, similar analytical models were performed to assess association between ANA and clinical and radiological characteristics. RESULTS: A final cohort of 258 patients was analyzed (out of 542 referrals): 106 nonMS and 152 with MS. There was no association between MS (vs. nonMS) diagnosis and ANA status (ANA positive n = 74) in the multivariable models (OR 1.5, 95 % CI 0.82, 2.72, p = 0.20). Among those with MS, there was no association of ANA seropositivity with the odds of atypical brain MRI features, number of cardinal MRI areas involved, location of MRI lesions, or of having an atypical presentation of first demyelinating event. Black race (OR 2.8, 95 % CI 1.27, 6.26, p = 0.01) and family history of autoimmunity (OR 2.1, 95 % CI 1.09, 3.98, p = 0.03) were independently associated with increased odds of ANA positivity. Within the MS cohort analysis, progressive MS (PMS; vs relapsing-remitting MS), a covariate in the model, appeared to be at higher odds of being ANA positive (OR 3.6, 95 % CI 1.03, 13.05, p = 0.046) but only when assessing mean area of cardinal MS locations. CONCLUSIONS: While ANA testing does not appear to be useful in distinguishing MS from non-MS, it remains less clear as to whether it may be associated with differences in the clinical course of MS (relapsing-remitting vs progressive). Future studies should aim to systematically evaluate whether those who are ANA positive are more likely, in well-designed and representative prospective cohorts, to be diagnosed with or develop progressive MS. Whether a positive ANA early in MS is associated with increased risk over time of developing or diagnosing another systemic autoimmune disease would also be of interest.

Dual-Energy Computed Tomography in Stroke Imaging : Value of a New Image Acquisition Technique for Ischemia Detection after Mechanical Thrombectomy.

OBJECTIVE: To assess if a new dual-energy computed tomography (DECT) technique enables an improved visualization of ischemic brain tissue after mechanical thrombectomy in acute stroke patients. MATERIAL AND METHODS: The DECT head scans with a new sequential technique (TwinSpiral DECT) were performed in 41 patients with ischemic stroke after endovascular thrombectomy and were retrospectively included. Standard mixed and virtual non-contrast (VNC) images were reconstructed. Infarct visibility and image noise were assessed qualitatively by two readers using a 4-point Likert scale. Quantitative Hounsfield units (HU) were used to assess density differences of ischemic brain tissue versus healthy tissue on the non-affected contralateral hemisphere. RESULTS: Infarct visibility was significantly better in VNC compared to mixed images for both readers R1 (VNC: median 1 (range 1-3), mixed: median 2 (range 1-4), p < 0.05) and R2 (VNC: median 2 (range 1-3), mixed: 2 (range 1-4), p < 0.05). Qualitative image noise was significantly higher in VNC compared to mixed images for both readers R1 (VNC: median 3, mixed: 2) and R2 (VNC: median 2, mixed: 1, p < 0.05, each). Mean HU were significantly different between the infarcted tissue and the reference healthy brain tissue on the contralateral hemisphere in VNC (infarct 24 ± 3) and mixed images (infarct 33 ± 5, p < 0.05, each). The mean HU difference between ischemia and reference in VNC images (mean 8 ± 3) was significantly higher (p < 0.05) compared to the mean HU difference in mixed images (mean 5 ± 4). CONCLUSION: TwinSpiral DECT allows an improved qualitative and quantitative visualization of ischemic brain tissue in ischemic stroke patients after endovascular treatment.

A Novel Dual-Energy CT Method for Detection and Differentiation of Intracerebral Hemorrhage From Contrast Extravasation in Stroke Patients After Endovascular Thrombectomy : Feasibility and First Results.

PURPOSE: Dual-energy computed tomography (DECT) has been shown to be able to differentiate between intracranial hemorrhage (ICH) and extravasation of iodinated contrast media (contrast staining [CS]). TwinSpiral DECT is a recently introduced technique, which allows image acquisition at two different energy levels in two consecutive spiral scans. The aim of this study was to evaluate the feasibility and accuracy of TwinSpiral DECT to distinguish between ICH and CS after endovascular thrombectomy (EVT) in patients with acute ischemic stroke. METHODS: This retrospective single-center study conducted between November 2019 and July 2020 included non-contrast TwinSpiral DECT scans (tube voltages 80 and 150Sn kVp) of 39 ischemic stroke patients (18 females, 21 males, mean age 69 ± 11 years) within 48-72 h after endovascular thrombectomy. Parenchymal hyperdensity was assessed for the presence of ICH or/and CS by two board certified and fellowship-trained, blinded and independent neuroradiologists using standard mixed images and virtual non-contrast (VNC) images with corresponding iodine maps from TwinSpiral DECT. Follow-up examinations (FU; CT or MRI) were used as a standard of reference. Sensitivity, specificity, and accuracy for the detection of ICH as well as the inter-reader agreement were calculated. RESULTS: Parenchymal hyperdensities were detected in 17/39 (44%) patients. Using DECT, they were classified by both readers as ICH in 9 (53%), CS in 8 (47%), and mixture of both in 6 (35%) cases with excellent agreement (κ = 0.81, P < 0.0001). The sensitivity, specificity, and accuracy for the detection of ICH in DECT was 90% (95% confidence interval [CI]: 84-96%), 100% (95% CI 94-100%) and 95% (95% CI 89-100%), and in mixed images 90% (95% CI 84-96%), 86% (95% CI 80-92%) and 88% (95% CI 82-94%), respectively. Inter-reader agreement for detecting ICH on DECT compared to the mixed images was κ = 1.00 (P < 0.0001) vs. κ = 0.51 (P = 0.034). CONCLUSION: TwinSpiral DECT demonstrates high accuracy and excellent specificity for differentiating ICH from CS in patients after mechanical thrombectomy due to acute ischemic stroke, and improves inter-reader agreement for detecting ICH compared to the standard mixed images.