Evaluating the efficacy of few-shot learning for GPT-4Vision in neurodegenerative disease histopathology: A comparative analysis with convolutional neural network model.

AIMS: Recent advances in artificial intelligence, particularly with large language models like GPT-4Vision (GPT-4V)-a derivative feature of ChatGPT-have expanded the potential for medical image interpretation. This study evaluates the accuracy of GPT-4V in image classification tasks of histopathological images and compares its performance with a traditional convolutional neural network (CNN). METHODS: We utilised 1520 images, including haematoxylin and eosin staining and tau immunohistochemistry, from patients with various neurodegenerative diseases, such as Alzheimer's disease (AD), progressive supranuclear palsy (PSP) and corticobasal degeneration (CBD). We assessed GPT-4V's performance using multi-step prompts to determine how textual context influences image interpretation. We also employed few-shot learning to enhance improvements in GPT-4V's diagnostic performance in classifying three specific tau lesions-astrocytic plaques, neuritic plaques and tufted astrocytes-and compared the outcomes with the CNN model YOLOv8. RESULTS: GPT-4V accurately recognised staining techniques and tissue origin but struggled with specific lesion identification. The interpretation of images was notably influenced by the provided textual context, which sometimes led to diagnostic inaccuracies. For instance, when presented with images of the motor cortex, the diagnosis shifted inappropriately from AD to CBD or PSP. However, few-shot learning markedly improved GPT-4V's diagnostic capabilities, enhancing accuracy from 40% in zero-shot learning to 90% with 20-shot learning, matching the performance of YOLOv8, which required 100-shot learning to achieve the same accuracy. CONCLUSIONS: Although GPT-4V faces challenges in independently interpreting histopathological images, few-shot learning significantly improves its performance. This approach is especially promising for neuropathology, where acquiring extensive labelled datasets is often challenging.

Upper motor neuron-predominant motor neuron disease presenting as atypical parkinsonism: A clinicopathological study.

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disorder characterized by upper and lower motor neuron signs. There are, however, cases where upper motor neurons (UMNs) are predominantly affected, leading to clinical presentations of UMN-dominant ALS or primary lateral sclerosis. Furthermore, cases exhibiting an UMN-predominant pattern of motor neuron disease (MND) presenting with corticobasal syndrome (CBS) have been sparsely reported. This study aims to clarify the clinicopathological features of patients with UMN-predominant MND. We reviewed 24 patients with UMN-predominant MND with TDP-43 pathology in the presence or absence of frontotemporal lobar degeneration. Additionally, we reviewed the medical records of patients with pathologically-confirmed corticobasal degeneration (CBD) who received a final clinical diagnosis of CBS (n = 10) and patients with pathologically-confirmed progressive supranuclear palsy (PSP) who received a final clinical diagnosis of PSP syndrome (n = 10). Of 24 UMN-predominant MND patients, 20 had a clinical diagnosis of an atypical parkinsonian disorder, including CBS (n = 11) and PSP syndrome (n = 8). Only two patients had antemortem diagnoses of motor neuron disease. UMN-predominant MND patients with CBS less frequently exhibited apraxia than those with CBD, and they were less likely to meet clinical criteria for possible or probable CBS. Similarly, UMN-predominant MND patients with PSP syndrome less often met clinical criteria for probable PSP than PSP patients with PSP syndrome. Our findings suggest that UMN-predominant MND can mimic atypical parkinsonism, and should be considered in the differential diagnosis of CBS and PSP syndrome, in particular when criteria are not met.

Gliovascular transcriptional perturbations in Alzheimer's disease reveal molecular mechanisms of blood brain barrier dysfunction.

To uncover molecular changes underlying blood-brain-barrier dysfunction in Alzheimer's disease, we performed single nucleus RNA sequencing in 24 Alzheimer's disease and control brains and focused on vascular and astrocyte clusters as main cell types of blood-brain-barrier gliovascular-unit. The majority of the vascular transcriptional changes were in pericytes. Of the vascular molecular targets predicted to interact with astrocytic ligands, SMAD3, upregulated in Alzheimer's disease pericytes, has the highest number of ligands including VEGFA, downregulated in Alzheimer's disease astrocytes. We validated these findings with external datasets comprising 4,730 pericyte and 150,664 astrocyte nuclei. Blood SMAD3 levels are associated with Alzheimer's disease-related neuroimaging outcomes. We determined inverse relationships between pericytic SMAD3 and astrocytic VEGFA in human iPSC and zebrafish models. Here, we detect vast transcriptome changes in Alzheimer's disease at the gliovascular-unit, prioritize perturbed pericytic SMAD3-astrocytic VEGFA interactions, and validate these in cross-species models to provide a molecular mechanism of blood-brain-barrier disintegrity in Alzheimer's disease.

Role of GBA variants in Lewy body disease neuropathology.

Rare and common GBA variants are risk factors for both Parkinson's disease (PD) and dementia with Lewy bodies (DLB). However, the degree to which GBA variants are associated with neuropathological features in Lewy body disease (LBD) is unknown. Herein, we assessed 943 LBD cases and examined associations of 15 different neuropathological outcomes with common and rare GBA variants. Neuropathological outcomes included LBD subtype, presence of a high likelihood of clinical DLB (per consensus guidelines), LB counts in five cortical regions, tyrosine hydroxylase immunoreactivity in the dorsolateral and ventromedial putamen, ventrolateral substantia nigra neuronal loss, Braak neurofibrillary tangle (NFT) stage, Thal amyloid phase, phospho-ubiquitin (pS65-Ub) level, TDP-43 pathology, and vascular disease. Sequencing of GBA exons revealed a total of 42 different variants (4 common [MAF > 0.5%], 38 rare [MAF < 0.5%]) in our series, and 165 cases (17.5%) had a copy of the minor allele for ≥ 1 variant. In analysis of common variants, p.L483P was associated with a lower Braak NFT stage (OR = 0.10, P < 0.001). In gene-burden analysis, presence of the minor allele for any GBA variant was associated with increased odds of a high likelihood of DLB (OR = 2.00, P < 0.001), a lower Braak NFT stage (OR = 0.48, P < 0.001), a lower Thal amyloid phase (OR = 0.55, P < 0.001), and a lower pS65-Ub level (ß: -0.37, P < 0.001). Subgroup analysis revealed that GBA variants were most common in LBD cases with a combination of transitional/diffuse LBD and Braak NFT stage 0-II or Thal amyloid phase 0-1, and correspondingly that the aforementioned associations of GBA gene-burden with a decreased Braak NFT stage and Thal amyloid phase were observed only in transitional or diffuse LBD cases. Our results indicate that in LBD, GBA variants occur most frequently in cases with greater LB pathology and low AD pathology, further informing disease-risk associations of GBA in PD, PD dementia, and DLB.

Cerebral artery overexpression of the NMUR1 gene is associated with moyamoya disease: a weighted gene co-expression network analysis.

INTRODUCTION: This study aimed to elucidate mechanisms underlying moyamoya disease (MMD) pathogenesis and to identify potential novel biomarkers. We utilized gene coexpression networks to identify hub genes associated with the disease. METHODS: Twenty-one middle cerebral artery (MCA) samples from MMD patients and 11 MCA control samples were obtained from the Gene Expression Omnibus (GEO) dataset, GSE189993. To discover functional pathways and potential biomarkers, weighted gene coexpression network analysis (WGCNA) was employed. The hub genes identified were re-assessed through differential gene expression analysis (DGEA) via DESeq2 for further reliability verification. An additional 4 samples from the superficial temporal arteries (STA) from MMD patients were obtained from GSE141025 and a subgroup analysis stratified by arterial type (MCA vs. STA) DGEA was performed to assess if the hub genes associated with MMD are expressed significantly greater on the affected arteries compared to healthy ones in MMD. RESULTS: WGCNA revealed a predominant module encompassing 139 hub genes, predominantly associated with the neuroactive ligand-receptor interaction (NLRI) pathway. Of those, 17 genes were validated as significantly differentially expressed. Neuromedin U receptor 1 (NMUR1) and thyrotropin-releasing hormone (TRH) were 2 out of the 17 hub genes involved in the NLRI pathway (log fold change [logFC]: 1.150, p = 0.00028; logFC: 1.146, p = 0.00115, respectively). MMD-only subgroup analysis stratified by location showed that NMUR1 is significantly overexpressed in the MCA compared to the STA (logFC: 1.962; p = 0.00053) which further suggests its possible localized involvement in the progressive stenosis seen in the cerebral arteries in MMD. CONCLUSION: This is the first study to have performed WGCNA on samples directly affected by MMD. NMUR1 expression is well known to induce localized arterial smooth muscle constriction and recently, type 2 inflammation which can predispose to arterial stenosis potentially advancing the symptoms and progression of MMD. Further validation and functional studies are necessary to understand the precise role of NMUR1 upregulation in MMD and its potential implications.

Senile plaque-associated transactive response DNA-binding protein 43 in Alzheimer's disease: A case report spanning 16 years of memory loss.

Transactive response DNA-binding protein 43 (TDP-43) pathological inclusions are found in frontotemporal lobar degeneration (FTLD-TDP) and Alzheimer's disease (AD-TDP). While clinically different, TDP-43 inclusions in FTLD-TDP and AD can have similar morphological characteristics. However, TDP-43 colocalizing with tau and forming "apple-bite" or "flame-shaped" neuronal cytoplasmic inclusions (NCI) are only found in AD-TDP. Here, we describe a case with AD and neuritic plaque-associated TDP-43. The patient was a 96-year-old right-handed Caucasian woman who had developed a slowly progressive amnestic syndrome compatible with typical AD at age 80. Genetic testing revealed APOE ε3/ε4, GRN r5848 CT, and MAPT H1/H2 genotype. Consistent with the old age at onset and long disease duration, limbic-predominant AD was found at autopsy, with high hippocampal yet low cortical neurofibrillary tangle (NFT) counts. Hippocampal and amygdala sclerosis were present. Immunohistochemistry for phospho-TDP-43 showed NCIs, dystrophic neurites, and rare neuronal intranuclear inclusions consistent with FTLD-TDP type A, as well as tau NFT-associated TDP-43 inclusions. These were frequent in the amygdala, entorhinal cortex, hippocampus, occipitotemporal gyrus, and inferior temporal gyrus but sparse in the mid-frontal cortex. Additionally, there were TDP-43-immunoreactive inclusions forming plaque-like structures in the molecular layer of the dentate fascia of the hippocampus. The presence of neuritic plaques in the same region was confirmed using thioflavin-S fluorescent microscopy and immunohistochemistry for phospho-tau. Double labeling immunofluorescence showed colocalization of TDP-43 and tau within neuritic plaques. Other pathologies included mild Lewy body pathology predominantly affecting the amygdala and olfactory bulb, aging-related tau astrogliopathy, and mixed small vessel disease (arteriolosclerosis and amyloid angiopathy) with several cortical microinfarcts. In conclusion, we have identified TDP-43 colocalizing with tau in neuritic plaques in AD, which expands the association of TDP-43 and tau in AD beyond NFTs. The clinical correlate of this plaque-associated TDP-43 appears to be a slowly progressive amnestic syndrome.

Evaluating the performance of large language models: ChatGPT and Google Bard in generating differential diagnoses in clinicopathological conferences of neurodegenerative disorders.

This study explores the utility of the large language models (LLMs), specifically ChatGPT and Google Bard, in predicting neuropathologic diagnoses from clinical summaries. A total of 25 cases of neurodegenerative disorders presented at Mayo Clinic brain bank Clinico-Pathological Conferences were analyzed. The LLMs provided multiple pathologic diagnoses and their rationales, which were compared with the final clinical diagnoses made by physicians. ChatGPT-3.5, ChatGPT-4, and Google Bard correctly made primary diagnoses in 32%, 52%, and 40% of cases, respectively, while correct diagnoses were included in 76%, 84%, and 76% of cases, respectively. These findings highlight the potential of artificial intelligence tools like ChatGPT in neuropathology, suggesting they may facilitate more comprehensive discussions in clinicopathological conferences.

Frequency of Comorbid Pathologies and Their Clinical Impact in Multiple System Atrophy.

BACKGROUND: Mixed pathology is common at autopsy for a number of age-associated neurodegenerative disorders; however, the frequency of comorbid pathologies in multiple system atrophy (MSA) and their clinical correlations are poorly understood. OBJECTIVE: We determined the frequency of comorbid pathologic processes in autopsy-confirmed MSA and assessed their clinical correlates. METHODS: This study included 160 neuropathologically established MSA from the Mayo Clinic brain bank. Clinical information, including age at onset or death, clinical subtype, initial symptoms, antemortem clinical diagnosis, and cognitive dysfunction was collected. We assessed comorbid pathologies including Alzheimer's disease neuropathologic change, Lewy-related pathology, argyrophilic grain disease, age-related τ astrogliopathy, transactive DNA-binding protein 43 pathology, cerebral amyloid angiopathy, and cerebrovascular small vessel disease and examined their clinical impact. RESULTS: The majority of MSA patients (62%) had no significant comorbid pathologies. There was a positive correlation between age at onset or death with the number of comorbid pathologies; however, even in the highest quartile group (average age at death 78 ± 6 years), the average number of comorbid pathologies was <2. Logistic regression analysis revealed that none of the assessed variables, including sex, age at onset, and the presence or absence of each comorbid pathology, were significantly associated with cognitive dysfunction. CONCLUSIONS: The majority of MSA patients do not have comorbid pathologies, even in advanced age, indicating that MSA is unique among neurodegenerative disorders in this regard. There was minimal clinical impact of comorbid pathologies in MSA. These findings warrant focusing on α-synuclein for the treatment strategy for MSA. © 2023 International Parkinson and Movement Disorder Society.

Molecular fragment characteristics and distribution of tangle associated TDP-43 (TATs) and other TDP-43 lesions in Alzheimer's disease.

TAR DNA binding protein 43 (TDP-43) pathology is a defining feature of frontotemporal lobar degeneration (FTLD). In FTLD-TDP there is a moderate-to-high burden of morphologically distinctive TDP-43 immunoreactive inclusions distributed throughout the brain. In Alzheimer's disease (AD), similar TDP-43 immunoreactive inclusions are observed. In AD, however, there is a unique phenomenon of neurofibrillary tangle-associated TDP-43 (TATs) whereby TDP-43 intermingles with neurofibrillary tangles. Little is known about the characteristics and distribution of TATs, or how burden and distribution of TATs compares to burden and distribution of other FTLD-TDP-like lesions observed in AD. Here we characterize molecular fragment characteristics, burden and distribution of TATs and assess how these features compare to features of other TDP-43 lesions. We performed TDP-43 immunohistochemistry with anti-phosphorylated, C- and N-terminal TDP-43 antibodies in 20 high-probability AD cases and semi-quantitative burden of seven inclusion types within five brain regions (entorhinal cortex, subiculum, CA1 and dentate gyrus of hippocampus, occipitotemporal cortex). Hierarchical cluster analysis was used to analyze the dataset that consisted of 75 different combinations of neuropathological features. TATs were nonspherical with heterogeneous staining patterns and present in all regions except hippocampal dentate. All three antibodies detected TATs although N-terminal antibody sensitivity was low. Three clusters were identified: Cluster-1 had mild-moderate TATs, moderate-frequent neuronal cytoplasmic inclusions, dystrophic neurites, neuronal intranuclear inclusions and fine neurites, and perivascular and granular inclusions identified only with the N-terminal antibody throughout the brain; Cluster-2 had scant TATs in limbic regions and Cluster-3 mild-moderate TATs and mild-moderate neuronal cytoplasmic inclusions and dystrophic neurites throughout the brain and moderate fine neurites. Only 17% of cluster 1 cases had the TMEM106b GG (protective) haplotype and 83% had hippocampal sclerosis. Both features differed across clusters (p=0.03 & p=0.01). TATs have molecular characteristics, distribution and burden, and genetic and pathologic associations like FTLD-TDP lesions.

Clinicopathologic features of a novel star-shaped transactive response DNA-binding protein 43 (TDP-43) pathology in the oldest old.

Transactive response DNA-binding protein 43 (TDP-43) pathology is categorized as type A-E in frontotemporal lobar degeneration and as type α-ß in Alzheimer disease (AD) based on inclusion type. We screened amygdala slides of 131 cases with varying ages at death, clinical/neuroimaging findings, and AD neuropathologic changes for TDP-43 pathology using anti-phospho-TDP-43 antibodies. Seven cases (5%) only showed atypical TDP-43 inclusions that could not be typed. Immunohistochemistry and immunofluorescence assessed the atypical star-shaped TDP-43 pathology including its distribution, species, cellular localization, and colocalization with tau. All 7 had died at an extremely old age (median: 100 years [IQR: 94-101]) from nonneurological causes and none had dementia (4 cognitively unimpaired, 3 with amnestic mild cognitive impairment). Neuroimaging showed mild medial temporal involvement. Pathologically, the star-shaped TDP-43-positive inclusions were found in medial (subpial) amygdala and, occasionally, in basolateral regions. Hippocampus only showed TDP-43-positive neurites in the fimbria and subiculum while the frontal lobe was free of TDP-43 inclusions. The star-shaped inclusions were better detected with antibodies against N-terminal than C-terminal TDP-43. Double-labeling studies confirmed deposition of TDP-43 within astrocytes and colocalization with tau. We have identified a novel TDP-43 pathology with star-shaped morphology associated with superaging, with a homogeneous clinicopathologic picture, possibly representing a novel, true aging-related TDP-43 pathology.