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.

Ultrasound evaluation of brain parenchyma in preterm infants with prenatal opioid exposure.

OBJECTIVE: To evaluate whether preterm infants with prenatal opioid exposure had differences in brain size on head ultrasounds (HUS) in comparison to non-exposed infants. STUDY DESIGN: Preterm infants ≤34 weeks with prenatal opioid exposure (n = 47) and matched non-exposed infants (n = 62) with early HUSs were examined. Fifteen brain measurements were made and linear regression models performed to evaluate differences. RESULTS: Brain measurements were smaller in the right ventricular index [ß = -0.18 mm (95% CI -0.32, -0.03]), left ventricular index [ß = -0.04 mm (95% CI -0.08, -0.003)], left basal ganglia insula [ß = -0.10 mm (95% CI -0.15, -0.04)], right basal ganglia insula [ß = -0.08 mm (95% CI -0.14, -0.03)], corpus callosum fastigium length [ß = -0.16 mm (95% CI -0.25, -0.06)], intracranial height index [ß = -0.31 mm (95% CI -0.44, -0.18)], and transcerebellar measurements [ß = -0.13 (95% CI -0.25, -0.02)] in the opioid-exposed group. CONCLUSIONS: Preterm infants with prenatal opioid exposure have smaller brain sizes compared to non-exposed infants, potentially increasing their risk for neurodevelopmental abnormalities.

Bayesian reconstruction of magnetic resonance images using Gaussian processes.

A central goal of modern magnetic resonance imaging (MRI) is to reduce the time required to produce high-quality images. Efforts have included hardware and software innovations such as parallel imaging, compressed sensing, and deep learning-based reconstruction. Here, we propose and demonstrate a Bayesian method to build statistical libraries of magnetic resonance (MR) images in k-space and use these libraries to identify optimal subsampling paths and reconstruction processes. Specifically, we compute a multivariate normal distribution based upon Gaussian processes using a publicly available library of T1-weighted images of healthy brains. We combine this library with physics-informed envelope functions to only retain meaningful correlations in k-space. This covariance function is then used to select a series of ring-shaped subsampling paths using Bayesian optimization such that they optimally explore space while remaining practically realizable in commercial MRI systems. Combining optimized subsampling paths found for a range of images, we compute a generalized sampling path that, when used for novel images, produces superlative structural similarity and error in comparison to previously reported reconstruction processes (i.e. 96.3% structural similarity and < 0.003 normalized mean squared error from sampling only 12.5% of the k-space data). Finally, we use this reconstruction process on pathological data without retraining to show that reconstructed images are clinically useful for stroke identification. Since the model trained on images of healthy brains could be directly used for predictions in pathological brains without retraining, it shows the inherent transferability of this approach and opens doors to its widespread use.

Vascular Collagen Type-IV in Hypertension and Cerebral Small Vessel Disease.

BACKGROUND: Cerebral small vessel disease (SVD) is common in older people and causes lacunar stroke and vascular cognitive impairment. Risk factors include old age, hypertension and variants in the genes COL4A1/COL4A2 encoding collagen alpha-1(IV) and alpha-2(IV), here termed collagen-IV, which are core components of the basement membrane. We tested the hypothesis that increased vascular collagen-IV associates with clinical hypertension and with SVD in older persons and with chronic hypertension in young and aged primates and genetically hypertensive rats. METHODS: We quantified vascular collagen-IV immunolabeling in small arteries in a cohort of older persons with minimal Alzheimer pathology (N=52; 21F/31M, age 82.8±6.95 years). We also studied archive tissue from young (age range 6.2-8.3 years) and older (17.0-22.7 years) primates (M mulatta) and compared chronically hypertensive animals (18 months aortic stenosis) with normotensives. We also compared genetically hypertensive and normotensive rats (aged 10-12 months). RESULTS: Collagen-IV immunolabeling in cerebral small arteries of older persons was negatively associated with radiological SVD severity (ρ: -0.427, P=0.005) but was not related to history of hypertension. General linear models confirmed the negative association of lower collagen-IV with radiological SVD (P<0.017), including age as a covariate and either clinical hypertension (P<0.030) or neuropathological SVD diagnosis (P<0.022) as fixed factors. Reduced vascular collagen-IV was accompanied by accumulation of fibrillar collagens (types I and III) as indicated by immunogold electron microscopy. In young and aged primates, brain collagen-IV was elevated in older normotensive relative to young normotensive animals (P=0.029) but was not associated with hypertension. Genetically hypertensive rats did not differ from normotensive rats in terms of arterial collagen-IV. CONCLUSIONS: Our cross-species data provide novel insight into sporadic SVD pathogenesis, supporting insufficient (rather than excessive) arterial collagen-IV in SVD, accompanied by matrix remodeling with elevated fibrillar collagen deposition. They also indicate that hypertension, a major risk factor for SVD, does not act by causing accumulation of brain vascular collagen-IV.

Structural MRI profiles and tau correlates of atrophy in autopsy-confirmed CTE.

BACKGROUND: Chronic traumatic encephalopathy (CTE), a neurodegenerative tauopathy, cannot currently be diagnosed during life. Atrophy patterns on magnetic resonance imaging could be an effective in vivo biomarker of CTE, but have not been characterized. Mechanisms of neurodegeneration in CTE are unknown. Here, we characterized macrostructural magnetic resonance imaging features of brain donors with autopsy-confirmed CTE. The association between hyperphosphorylated tau (p-tau) and atrophy on magnetic resonance imaging was examined. METHODS: Magnetic resonance imaging scans were obtained by medical record requests for 55 deceased symptomatic men with autopsy-confirmed CTE and 31 men (n = 11 deceased) with normal cognition at the time of the scan, all >60 years Three neuroradiologists visually rated regional atrophy and microvascular disease (0 [none]-4 [severe]), microbleeds, and cavum septum pellucidum presence. Neuropathologists rated tau severity and atrophy at autopsy using semi-quantitative scales. RESULTS: Compared to unimpaired males, donors with CTE (45/55=stage III/IV) had greater atrophy of the orbital-frontal (mean diff.=1.29), dorsolateral frontal (mean diff.=1.31), superior frontal (mean diff.=1.05), anterior temporal (mean diff.=1.57), and medial temporal lobes (mean diff.=1.60), and larger lateral (mean diff.=1.72) and third (mean diff.=0.80) ventricles, controlling for age at scan (ps<0.05). There were no effects for posterior atrophy or microvascular disease. Donors with CTE had increased odds of a cavum septum pellucidum (OR = 6.7, p < 0.05). Among donors with CTE, greater tau severity across 14 regions corresponded to greater atrophy on magnetic resonance imaging (beta = 0.68, p < 0.01). CONCLUSIONS: These findings support frontal-temporal atrophy as a magnetic resonance imaging finding of CTE and show p-tau accumulation is associated with atrophy in CTE.

The relationship of age and hypertension with cognition and gray matter cerebral blood volume in a rhesus monkey model of human aging.

The goal of this study was to investigate whether alterations in cerebral microvasculature, as measured by cerebral blood volume (CBV), contribute to age- and hypertension-related impairments in cognitive function with a focus on executive function and memory. Data were collected on 19 male rhesus monkeys ranging from 6.4 to 21.6 years of age. Hypertension was induced through surgical coarctation of the thoracic aorta. We assessed whether performance on tasks of memory and executive function corresponded to CBV in either the hippocampus or prefrontal cortex. We found a relationship between duration of hypertension and CBV in the gray matter of the prefrontal cortex, but not the hippocampus. No relationships were found with the degree of hypertension or age. Increased prefrontal CBV was related to greater impairment in executive function while hippocampal CBV was not related to memory performance. These findings suggest that duration, but not severity, of hypertension or age are important factors underlying alterations in brain microvasculature and that executive function is more vulnerable than memory function. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Implementation of Institutional Triaging Algorithms Decreases Head and Neck MDCT Use in Blunt Trauma.

Background Multidetector CT (MDCT) enables rapid and accurate diagnosis of head and neck (HN) injuries in patients with blunt trauma (BT). However, MDCT is overused, and appropriate selection of patients for imaging could improve workflow. Purpose To investigate the effect of implementing clinical triaging algorithms on use of MDCT in the HN in patients who have sustained BT. Materials and Methods In this retrospective study, patients aged 15 years or older with BT admitted between October 28, 2007, and December 31, 2013, were included. Patients were divided into pre- and postalgorithm groups. The institutional trauma registry and picture archiving and communication system reports were reviewed to determine which patients underwent MDCT of the head, MDCT of the cervical spine (CS), and MDCT angiography of the HN at admission and whether these examinations yielded positive results. Injury Severity Score, Acute Physiology and Chronic Health Evaluation II score (only those patients in the intensive care unit), length of hospital stay (LOS), length of intensive care unit stay (ICULOS), and mortality were obtained from the trauma registry. Results A total of 8999 patients (mean age, 45 years ± 20 [standard deviation]; age range, 15-101 years; 6027 male) were included in this study. A lower percentage of the postalgorithm group versus the prealgorithm group underwent MDCT of the head (55.8% [2774 of 4969 patients]; 95% CI: 54.4, 57.2 vs 64.2% [2589 of 4030 patients]; 95% CI: 62.8, 65.7; P < .001) and CS (49.4% [2452 of 4969 patients]; 95% CI: 48.0, 50.7 vs 60.5% [2438 of 4030 patients]; 95% CI: 59.0, 62.0; P < .001) but not MDCT angiography of the HN (9.7% [480 of 4969 patients]; 95% CI: 8.9, 10.5 vs 9.8% [393 of 4030 patients]; 95% CI: 8.9, 10.7; P > .99). Pre- versus postalgorithm groups did not differ in LOS (mean, 4.8 days ± 7.1 vs 4.5 days ± 7.1, respectively; P = .42), ICULOS (mean, 4.6 days ± 6.6 vs 4.8 days ± 6.7, respectively; P > .99), or mortality (2.9% [118 of 4030 patients]; 95% CI: 2.5, 3.5; vs 2.8% [141 of 4969 patients]; 95% CI: 2.4, 3.3; respectively; P > .99). Conclusion Implementation of a clinical triaging algorithm resulted in decreased use of multidetector CT of the head and cervical spine in patients who experienced blunt trauma, without increased adverse outcomes. © RSNA, 2021 See also the editorial by Munera and Martin in this issue.

Nonhomogeneous Gadolinium Retention in the Cerebral Cortex after Intravenous Administration of Gadolinium-based Contrast Agent in Rats and Humans.

Background Gadolinium retention after repeated gadolinium-based contrast agent (GBCA) exposure has been reported in subcortical gray matter. However, gadolinium retention in the cerebral cortex has not been systematically investigated. Purpose To determine whether and where gadolinium is retained in rat and human cerebral cortex. Materials and Methods The cerebral cortex in Sprague-Dawley rats treated with gadopentetate dimeglumine (three doses over 4 weeks; cumulative gadolinium dose, 7.2 mmol per kilogram of body weight; n = 6) or saline (n = 6) was examined with antemortem MRI. Two human donors with repeated GBCA exposure (three and 15 doses; 1 and 5 months after exposure), including gadopentetate dimeglumine, and two GBCA-naive donors were also evaluated. Elemental brain maps (gadolinium, phosphorus, zinc, copper, iron) for rat and human brains were constructed by using laser ablation inductively coupled plasma mass spectrometry. Results Gadopentetate dimeglumine-treated rats showed region-, subregion-, and layer-specific gadolinium retention in the neocortex (anterior cingulate cortex: mean gadolinium concentration, 0.28 µg ∙ g-1 ± 0.04 [standard error of the mean]) that was comparable (P > .05) to retention in the allocortex (mean gadolinium concentration, 0.33 µg ∙ g-1 ± 0.04 in piriform cortex, 0.24 µg ∙ g-1 ± 0.04 in dentate gyrus, 0.17 µg ∙ g-1 ± 0.04 in hippocampus) and subcortical structures (0.47 µg ∙ g-1 ± 0.10 in facial nucleus, 0.39 µg ∙ g-1 ± 0.10 in choroid plexus, 0.29 µg ∙ g-1 ± 0.05 in caudate-putamen, 0.26 µg ∙ g-1 ± 0.05 in reticular nucleus of the thalamus, 0.24 µg ∙ g-1 ± 0.04 in vestibular nucleus) and significantly greater than that in the cerebellum (0.17 µg ∙ g-1 ± 0.03, P = .01) and white matter tracts (anterior commissure: 0.05 µg ∙ g-1 ± 0.01, P = .002; corpus callosum: 0.05 µg ∙ g-1 ± 0.02, P = .001; cranial nerve: 0.02 µg ∙ g-1 ± 0.01, P = .004). Retained gadolinium colocalized with parenchymal iron. T1-weighted MRI signal intensification was not observed. Gadolinium retention was detected in the cerebral cortex, pia mater, and pia-ensheathed leptomeningeal vessels in two GBCA-exposed human brains but not in two GBCA-naive human brains. Conclusion Repeated gadopentetate dimeglumine exposure is associated with gadolinium retention in specific regions, subregions, and layers of cerebral cortex that are critical for higher cognition, affect, and behavior regulation, sensorimotor coordination, and executive function. © RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Kanal in this issue.

Hippocampal network connections account for differences in memory performance in the middle-aged rhesus monkey.

Recent neurophysiological and functional neuroimaging studies suggest that the memory decline found with normal aging is not solely due to regional disruptions in the hippocampus, but also is brought about by alterations in the functional coupling between the hippocampus and long-distance neocortical regions. However, the anatomical basis for this functional "dyscoupling" has not been fully revealed. In this study, we applied a multimodal magnetic resonance imaging technique to noninvasively examine the large-scale anatomical and functional hippocampal network of a group of middle aged rhesus monkeys. Using diffusion spectrum imaging, we have found that monkeys with lower memory performance had weaker structural white matter connections between the hippocampus and neocortical association areas. Resting state functional imaging revealed somewhat of an opposite result. Monkeys with low memory performance displayed elevated coupling strengths in the network between the hippocampus and the neocortical areas. Taken together with recent findings, this contradictory pattern may be the result of either underlying physiological burden or abnormal neuronal decoupling due to the structural alterations, which induce a neuronal compensation mechanism for the structural loss or interference on task related neuronal activation, respectively.