AI-based differential diagnosis of dementia etiologies on multimodal data.

Differential diagnosis of dementia remains a challenge in neurology due to symptom overlap across etiologies, yet it is crucial for formulating early, personalized management strategies. Here, we present an artificial intelligence (AI) model that harnesses a broad array of data, including demographics, individual and family medical history, medication use, neuropsychological assessments, functional evaluations and multimodal neuroimaging, to identify the etiologies contributing to dementia in individuals. The study, drawing on 51,269 participants across 9 independent, geographically diverse datasets, facilitated the identification of 10 distinct dementia etiologies. It aligns diagnoses with similar management strategies, ensuring robust predictions even with incomplete data. Our model achieved a microaveraged area under the receiver operating characteristic curve (AUROC) of 0.94 in classifying individuals with normal cognition, mild cognitive impairment and dementia. Also, the microaveraged AUROC was 0.96 in differentiating the dementia etiologies. Our model demonstrated proficiency in addressing mixed dementia cases, with a mean AUROC of 0.78 for two co-occurring pathologies. In a randomly selected subset of 100 cases, the AUROC of neurologist assessments augmented by our AI model exceeded neurologist-only evaluations by 26.25%. Furthermore, our model predictions aligned with biomarker evidence and its associations with different proteinopathies were substantiated through postmortem findings. Our framework has the potential to be integrated as a screening tool for dementia in clinical settings and drug trials. Further prospective studies are needed to confirm its ability to improve patient care.

AI-based differential diagnosis of dementia etiologies on multimodal data.

Differential diagnosis of dementia remains a challenge in neurology due to symptom overlap across etiologies, yet it is crucial for formulating early, personalized management strategies. Here, we present an AI model that harnesses a broad array of data, including demographics, individual and family medical history, medication use, neuropsychological assessments, functional evaluations, and multimodal neuroimaging, to identify the etiologies contributing to dementia in individuals. The study, drawing on 51,269 participants across 9 independent, geographically diverse datasets, facilitated the identification of 10 distinct dementia etiologies. It aligns diagnoses with similar management strategies, ensuring robust predictions even with incomplete data. Our model achieved a micro-averaged area under the receiver operating characteristic curve (AUROC) of 0.94 in classifying individuals with normal cognition, mild cognitive impairment and dementia. Also, the micro-averaged AUROC was 0.96 in differentiating the dementia etiologies. Our model demonstrated proficiency in addressing mixed dementia cases, with a mean AUROC of 0.78 for two co-occurring pathologies. In a randomly selected subset of 100 cases, the AUROC of neurologist assessments augmented by our AI model exceeded neurologist-only evaluations by 26.25%. Furthermore, our model predictions aligned with biomarker evidence and its associations with different proteinopathies were substantiated through postmortem findings. Our framework has the potential to be integrated as a screening tool for dementia in various clinical settings and drug trials, with promising implications for person-level management.

Hypertrophic pachymeningitis of the internal auditory canal and cerebellopontine angle from a large recurrent attic cholesteatoma.

OBJECTIVES: To describe a rare complication of cholesteatoma. METHODS: Case report with literature review. RESULTS: We report a case of a 37-year-old male who presented for evaluation of otorrhea, headache, and progressive left sensorineural hearing loss. Clinical and radiologic evaluation demonstrated a large recurrent attic cholesteatoma with erosion into the lateral and superior semicircular canals, and diffuse enhancement of the internal auditory canal and cerebellopontine angle suggestive of hypertrophic pachymeningitis secondary to cholesteatoma. After treatment with a course of antibiotics and canal wall down mastoidectomy surgery for cholesteatoma exteriorization, he experienced improvement of his symptoms and resolution of hypertrophic pachymeningitis. CONCLUSION: Hypertrophic pachymeningitis is a rarely described complication of cholesteatoma. In the context of cholesteatoma, treatment with antibiotics and surgical removal or exteriorization of cholesteatoma are effective treatments for HP.

Real-World Performance of Large Vessel Occlusion Artificial Intelligence-Based Computer-Aided Triage and Notification Algorithms-What the Stroke Team Needs to Know.

PURPOSE: To evaluate the real-world performance of two FDA-approved artificial intelligence (AI)-based computer-aided triage and notification (CADt) detection devices and compare them with the manufacturer-reported performance testing in the instructions for use. MATERIALS AND METHODS: Clinical performance of two FDA-cleared CADt large-vessel occlusion (LVO) devices was retrospectively evaluated at two separate stroke centers. Consecutive "code stroke" CT angiography examinations were included and assessed for patient demographics, scanner manufacturer, presence or absence of CADt result, CADt result, and LVO in the internal carotid artery (ICA), horizontal middle cerebral artery (MCA) segment (M1), Sylvian MCA segments after the bifurcation (M2), precommunicating part of cerebral artery, postcommunicating part of the cerebral artery, vertebral artery, basilar artery vessel segments. The original radiology report served as the reference standard, and a study radiologist extracted the above data elements from the imaging examination and radiology report. RESULTS: At hospital A, the CADt algorithm manufacturer reports assessment of intracranial ICA and MCA with sensitivity of 97% and specificity of 95.6%. Real-world performance of 704 cases included 79 in which no CADt result was available. Sensitivity and specificity in ICA and M1 segments were 85.3% and 91.9%. Sensitivity decreased to 68.5% when M2 segments were included and to 59.9% when all proximal vessel segments were included. At hospital B the CADt algorithm manufacturer reports sensitivity of 87.8% and specificity of 89.6%, without specifying the vessel segments. Real-world performance of 642 cases included 20 cases in which no CADt result was available. Sensitivity and specificity in ICA and M1 segments were 90.7% and 97.9%. Sensitivity decreased to 76.4% when M2 segments were included and to 59.4% when all proximal vessel segments are included. DISCUSSION: Real-world testing of two CADt LVO detection algorithms identified gaps in the detection and communication of potentially treatable LVOs when considering vessels beyond the intracranial ICA and M1 segments and in cases with absent and uninterpretable data.

Deep learning for risk-based stratification of cognitively impaired individuals.

Quantifying the risk of progression to Alzheimer's disease (AD) could help identify persons who could benefit from early interventions. We used data from the Alzheimer's Disease Neuroimaging Initiative (ADNI, n = 544, discovery cohort) and the National Alzheimer's Coordinating Center (NACC, n = 508, validation cohort), subdividing individuals with mild cognitive impairment (MCI) into risk groups based on cerebrospinal fluid amyloid-ß levels and identifying differential gray matter patterns. We then created models that fused neural networks with survival analysis, trained using non-parcellated T1-weighted brain MRIs from ADNI data, to predict the trajectories of MCI to AD conversion within the NACC cohort (integrated Brier score: 0.192 [discovery], and 0.108 [validation]). Using modern interpretability techniques, we verified that regions important for model prediction are classically associated with AD. We confirmed AD diagnosis labels using postmortem data. We conclude that our framework provides a strategy for risk-based stratification of individuals with MCI and for identifying regions key for disease prognosis.

Revisiting the Tentorial Venous Sinuses: Anatomical and Histological Study.

BACKGROUND: Anatomical studies of the tentorial sinuses (TS) are scant, and to our knowledge, histological studies of this structure have not been reported. Therefore, we aim to better elucidate this anatomy. METHODS: In 15 fresh frozen, latex injected, adult cadaveric specimens, the TS were evaluated with microsurgical dissection and histology. RESULTS: The superior layer had a mean thickness of 0.22 mm, and the inferior layer had a mean thickness of 0.26 mm. Two types of TS were identified. Type 1 was a small intrinsic plexiform sinus with no obvious connections to the draining veins with gross examination. Type 2 was a larger tentorial sinus with direct connections to the bridging veins from the cerebral and cerebellar hemispheres. In general, type 1 sinuses were located more medially than type 2 sinuses. The inferior tentorial bridging veins drained directly into the TS along with connections to the straight and transverse sinuses. In 53.3% of specimens, superficial and deep sinuses were seen, with superior and inferior groups draining the cerebrum and cerebellum, respectively. CONCLUSIONS: We identified novel findings for the TS which can be considered surgically and when diagnosing pathology involves these venous sinuses.

Multimodal deep learning for Alzheimer's disease dementia assessment.

Worldwide, there are nearly 10 million new cases of dementia annually, of which Alzheimer's disease (AD) is the most common. New measures are needed to improve the diagnosis of individuals with cognitive impairment due to various etiologies. Here, we report a deep learning framework that accomplishes multiple diagnostic steps in successive fashion to identify persons with normal cognition (NC), mild cognitive impairment (MCI), AD, and non-AD dementias (nADD). We demonstrate a range of models capable of accepting flexible combinations of routinely collected clinical information, including demographics, medical history, neuropsychological testing, neuroimaging, and functional assessments. We then show that these frameworks compare favorably with the diagnostic accuracy of practicing neurologists and neuroradiologists. Lastly, we apply interpretability methods in computer vision to show that disease-specific patterns detected by our models track distinct patterns of degenerative changes throughout the brain and correspond closely with the presence of neuropathological lesions on autopsy. Our work demonstrates methodologies for validating computational predictions with established standards of medical diagnosis.

Basilar Artery Bands: Anatomic and Histologic Study with Application to Coiling and Stenting Procedures.

BACKGROUND: Internal bands of the basilar artery (BA) have been rarely studied. Because bands could have obvious consequences in patient diagnosis and care, the present multiinstitutional cadaveric study was performed. METHODS: The intraluminal bands of BAs were studied in 80 cadaveric specimens derived from India (Group 1) and the United States (Group 2). Their orientation within the vessel was recorded. Measurements included the length of the BA, diameter of the BA, intraluminal band length and thickness, and distance of the band to the vertebrobasilar junction. Selected bands were submitted for histologic analysis. RESULTS: In Group 1, an intraluminal band (all vertically oriented) was identified in 16.6%. The mean length and thickness of the bands were 2.04 mm and 1.33 mm, respectively. These bands were located at a mean distance of 1.74 mm superior to the vertebrobasilar junction. In Group 2, an intraluminal band was identified in 6%. One band was vertically positioned, and 2 were horizontally positioned. The mean length and thickness of the bands were 2.5 mm and 0.9 mm, respectively. These latter bands were located at a mean distance of 2.23 mm superior to the vertebrobasilar junction. Histologically, the bands were essentially extensions of the tunica media and interna of the artery. CONCLUSIONS: To our knowledge, this study is the first multiethnic study of the prevalence and morphometry of the BA bands. A better understanding of these bands may help reveal their relationship to thrombus and aneurysmal formation and their impact on endovascular procedures.

Diagnostic Features of Emphysematous Osteomyelitis.

Emphysematous osteomyelitis (EO) is a rare, aggressive, and potentially fatal variant of osteomyelitis related to gas-forming organisms. Imaging plays a vital role in diagnosis. The purpose of this study was to describe a novel and distinct imaging sign of EO, by analysis of the imaging characteristics of 3 newly identified cases of EO as well as all reported cases in the literature. Literature review and retrospective study in 2 tertiary care medical centers was conducted. During the course of clinical care over the last year, we have observed 2 cases of EO. An Institutional Review Board approved 1-year systematic retrospective review of our institutional radiology information system identified 1 additional case of EO. We conducted a MEDLINE literature search to identify all published EO case reports using key phrases. The imaging in all published cases was reviewed. To address the specificity of imaging characteristics, we identified cases demonstrating variants of intraosseous gas that were not caused by EO. We found 31 individual case reports of EO via MEDLINE search. Of the published cases, 25 of the 31 cases contained images of sufficient extent and quality for image analysis. Including our 3 cases, a total of 34 cases were identified. Of the 34 cases, 28 had images of sufficient quality and extent to determine the presence or absence of the "pumice stone" pattern. The "pumice stone" sign was identified in 27 of 28 cases (96%). The most commonly affected sites of infection include the pelvic bones (38%; 13 of 34), vertebral bodies (32%; 11 of 34), and femurs (24%; 8 of 34). Adjacent soft tissue inflammation and emphysema was evident in 23 of 29 cases (79%). Cortical destruction, a key diagnostic feature of traditional osteomyelitis, was absent in 79% of cases of EO. We describe the "pumice stone sign" as a radiological sign to identify EO. We found it present in 96% of all currently known cases of EO. EO commonly involves the pelvic bones (38%), vertebral bodies (32%), and femur (24%). Key features also include adjacent soft tissue emphysema (79%) and absence of cortical destruction (79%).

Acute Myelopathy: Vascular and Infectious Diseases.

Vascular and infectious causes are rare but important causes of spinal cord injury. High suspicion for these processes is necessary, as symptoms may progress over hours to days, resulting in delayed presentation and diagnosis and worse outcomes. History and clinical examination findings can assist with localization of the affected vascular territory and spinal level, which will assist with focusing spinal imaging. Open and/or endovascular surgical management depends on the associated vascular abnormality. Infectious myelopathy treatment consists of targeted antimicrobial therapy when possible, infectious source control, and again, close monitoring for systemic complications.

Spinal Epidural Venous Plexus Pathology in Hirayama Disease.

Hirayama disease is a rare juvenile amyotrophy that is often misdiagnosed as an unrelated, relentlessly progressive disease. We present the case of an 18-year-old man who presented with weakness and atrophy of the right forearm and hand. Dynamic cervical magnetic resonance imaging was used, revealing the classic findings of epidural venous plexus dilation and anterior displacement of the dural sac. In addition, dilation of the external vertebral venous plexus was visualized. We discuss the clinical utility of dynamic magnetic resonance imaging and the underlying pathophysiology of these findings in Hirayama disease.

CT Attenuation of Acute Subdural Hematomas in Patients with Anemia.

BACKGROUND AND PURPOSE: Isodense and hypodense acute subdural hematomas have been reported in the literature in anemic patients. The purpose of this study is to see if there is a statistically significant difference between the Hounsfield unit measurements of acute subdural hematomas in anemic and nonanemic patients. METHODS: A total of 109 patients were analyzed. We measured the most hyperdense component of the subdural hematoma and compared these measurements for both anemic and nonanemic patients. RESULTS: All patients with anemia had a hyperdense component to their subdural hematomas during the acute period. No statistically significant difference was found in the density of the subdural hematomas between the two groups. More heterogeneous subdural hematomas were found in the anemic group than the nonanemic group, which suggests that anemia alone is not a sufficient explanation for acute homogenous isodense and hypodense subdural hematomas. CONCLUSION: A hyperdense subdural component was present in all acute subdural hematomas in anemic patients. Therefore, anemia alone is not a sufficient explanation for a homogenous low-density acute subdural hematoma.

Radiation-Induced Painful Neurogenic Hypertrophy Treated With Botulinum Toxin A.

Painful neurogenic hypertrophy is a rare complication of radiation therapy. We report a 27-year-old woman with a history of adenoid cystic carcinoma of the submandibular gland presented with painful twitching of her left shoulder. Electrodiagnostic studies were consistent with a diagnosis of radiation-induced spinal accessory nerve hyperactivity. The patient failed conventional medical therapy. She was treated with an injection of botulinum toxin A, and within 1 month experienced significant relief of symptoms. We thus conclude that Botulinum toxin is a therapeutic option for the pain associated with radiation-induced peripheral nerve hyperactivity.

Usage of an Electronic Database and Checklist System for Improvement in Magnetic Resonance Imaging Acquisition.

PURPOSE: To determine whether implementation of an easily accessible electronic database promotes significant reporting of magnetic resonance imaging (MRI) acquisition errors. Additionally, we wanted to see if analysis of the error reports could be used to create a comprehensive checklist to avoid the most common errors. METHODS: A new, simple, and efficient electronic database reporting system was written in-house and implemented at our institution. Over the course of 4 months, the use of this database enabled collection and analysis of sufficient data for trend analysis. A simple 4-point checklist for MRI technologist use was developed based on the most commonly reported errors. Reported MRI acquisition error rates were collected and analyzed thereafter. RESULTS: By the first full month of implementation, MRI scan error reporting increased from a previous negligible baseline rate to 3.03%. The comprehensive checklist was based on the 4 most common issues reported. Verification of checklist use showed that adherence to this requirement averaged greater than 94%. Immediately following roll out of the checklist, the percentage of errors reported fell to 1.7% with a continued decline in error reports thereafter. An approximately 60% reduction in errors in the last month of the study was evident as compared to the first month of data collection. CONCLUSIONS: The use of an efficient error reporting system and implementation of a checklist based on the most common MRI acquisition errors results in a substantial decrease in the baseline MRI acquisition error rates.

Emergency Magnetic Resonance Imaging 3-Tiered Prioritization.

BACKGROUND/PURPOSE: Traditional methods for prioritization are limited and insufficient for today's magnetic resonance imaging (MRI) demands. In particular, the discrepancy in urgency of the heterogeneous emergency department (ED) patient population necessitates risk stratification to meet different degrees of urgency. The purpose of this study is to more effectively prioritize the MRI imaging needs of ED patients commensurate with the severity of their presenting illness. METHODS: A 3-level tiered classification system (tier 1: critical, tier 2: emergent, and tier 3: urgent) of ED patients with unambiguous hierarchically defined numerical classifications was implemented to replace a traditional method of MRI orders. Each tier was accompanied by guiding consensus-driven clinical definitions and common qualifying examples. Lastly, each tier imaging order was tied to a specific target "order to imaging start time" (OTST). After implementation, a month-by-month 1-year retrospective analysis of ED MRI imaging order volume was conducted to assess the percentage distribution of each category. In addition, a month-by-month 1-year retrospective analysis of the OTST for each tier was conducted. The OTST outcome measure was used to monitor the ability of the system to meet tier target times based on severity. RESULTS: The system effectively prioritized ED patients into 3 tiers based on acuity. An inverse relationship existed between ED MRI OTST and the tier severity into which the patient was stratified. We found that only 4% of the ED-specific volume is truly critical (tier 1). In addition, tier 3 MRI examinations constituted 75% of the ED volume. Month-by-month quality assurance analysis demonstrated consistent completion of examinations under or close to the target times tied to each tier. The average overall wait time from order time to begin scan time for all ED MRIs decreased from 245 minutes (4.1 hours) at baseline to less than 136 minutes (2.7 hours). CONCLUSIONS: We implemented and evaluated a 3-tiered system of ED MRI imaging orders based on patient severity. The system was unambiguous due to its numerical hierarchy, and each of the 3 tiers was accompanied by explicit guiding definitions for each category. A quality assurance process following implementation allowed us to monitor the ability of the system to meet target times tied to each tier. Our current ability to accurately predict a target performance time allows us to set accurate expectations for both providers and patients.

Case Report: Zika Virus Meningoencephalitis and Myelitis and Associated Magnetic Resonance Imaging Findings.

Zika virus (ZIKV) has a wide clinical spectrum of associated neurologic disease including microcephaly and Guillain-Barre syndrome but, despite its known neurotropism, ZIKV meningoencephalitis and myelitis have been rare complications. We describe a case of ZIKV meningoencephalitis and probable myelitis and its associated magnetic resonance imaging findings that rapidly resolved during recovery in a previously healthy adult.

Barraquer-Simons Syndrome.

BACKGROUND: Barraquer-Simons syndrome (BSS) is a rare acquired lipodystrophy characterized by gradually symmetric subcutaneous fat loss in a craniocaudal distribution, associated with hypocomplementemia, diabetes and hypertriglyceridemia. Few investigators have studied body fat distribution with cross-sectional imaging techniques. METHODS: We present 2 cases of BSS with emphasis on phenotypic analysis through cross-sectional imaging. RESULTS: For the first time, we demonstrate bone marrow involvement and deep cervical and axillary fat sparing of Barraquer-Simons using magnetic resonance imaging. CONCLUSION: Phenotypic analysis in lipodystrophies such as Barraquer-Simons is an essential guide for future experiments. Therefore, careful analysis of cross-sectional imaging should be conducted in future studies as areas of involvement or fat sparing may be overlooked. The major contributions of our work are that this is the first time that deep cervical or nuchal and axillary fat sparing and bone marrow involvement has been documented in BSS.