Machine Learning of Endoscopy Images to Identify, Classify, and Segment Sinonasal Masses

ABSTRACT Background We developed and assessed the performance of a machine learning model (MLM) to identify, classify, and segment sinonasal masses based on endoscopic appearance. Methods A convolutional neural network‐based model was constructed from nasal endoscopy images from patients evaluated at an otolaryngology center between 2013 and 2024. Images were classified into four groups: normal endoscopy, nasal polyps, benign, and malignant tumors. Polyps and tumors were confirmed with histopathological diagnosis. Images were annotated by an otolaryngologist and independently verified by two other otolaryngologists. We used high‐ and low‐quality images to mirror real‐world conditions. The models used for classification (EfficientNet‐B2) and segmentation (nnUNet) were trained, validated, and tested at an 8:1:1 ratio. The performance accuracy was averaged across a 10‐fold cross‐validation assessment. Segmentation accuracy was assessed via Dice similarity coefficients. Results A total of 1242 images from 311 patients were used. The MLM was trained, validated, and tested on 663 normal, 276 polyps, 157 benign, and 146 malignant tumors images. Overall, the model performed at 84.1 ± 4.3% accuracy in the validation set and 80.4 ± 1.7% in the test set. The model correctly identified the presence of a sinonasal mass at 90.5 ± 1.2% accuracy rate. The MLM accuracy performance rates were 86.2 ± 1.0% for polyps and 84.1 ± 1.8% for tumors. Benign and malignant tumor subclassification achieved 87.8 ± 2.1% and 94.0 ± 2.4% accuracy, respectively. Segmentation accuracies for polyps were 72.3% and 72.8% for tumors. Conclusions An MLM for nasal endoscopy images can perform with moderate to high accuracy in identifying, classifying, and segmenting sinonasal masses. Performance in future iterations may improve with larger and more diverse training datasets.

Olfactory Training: Recommendation Frequency amongst Rhinologists

Response to the Editor Regarding: “In Office Sampling of Eosinophil Peroxidase to Diagnose Eosinophilic Chronic Rhinosinusitis”

Differences in baseline cognitive performance between participants with early‐onset and late‐onset Alzheimer's disease: Comparison of LEADS and ADNI

Abstract INTRODUCTION Early‐onset Alzheimer's disease (EOAD) and late‐onset Alzheimer's disease (LOAD) share similar amyloid etiology, but evidence from smaller‐scale studies suggests that they manifest differently clinically. Current analyses sought to contrast the cognitive profiles of EOAD and LOAD. METHODS Z‐ score cognitive‐domain composites for 311 amyloid‐positive sporadic EOAD and 314 amyloid‐positive LOAD participants were calculated from baseline data from age‐appropriate control cohorts. Z‐ score composites were compared between AD groups for each domain. RESULTS After controlling for cognitive status, EOAD displayed worse visuospatial, executive functioning, and processing speed/attention skills relative to LOAD, and LOAD displayed worse language, episodic immediate memory, and episodic delayed memory. DISCUSSION Sporadic EOAD possesses distinct cognitive profiles relative to LOAD. Clinicians should be alert for non‐amnestic impairments in younger patients to ensure proper identification and intervention using disease‐modifying treatments. Highlights Both early‐onset Alzheimer's disease (EOAD) and late‐onset Alzheimer's disease (LOAD) participants displayed widespread cognitive impairments relative to their same‐aged peers. Cognitive impairments were more severe for EOAD than for LOAD participants in visuospatial and executive domains. Memory and language impairments were more severe for LOAD than for EOAD participants Results were comparable after removing clinical phenotypes of posterior cortical atrophy (PCA), primary progressive aphasia (lv‐PPA), and frontal‐variant AD.

The Consortium for Clarity in ADRD Research Through Imaging (CLARiTI)

Abstract The presence of multiple pathologies is the largest predictor of dementia. A major gap in the field is the in vivo detection of mixed pathologies and their antecedents. The Alzheimer's Disease Research Centers (ADRCs) are uniquely positioned to address this gap. The ADRCs longitudinally follow ≈ 17,000 participants, ranging from cognitively unimpaired to dementia, arising from Alzheimer's disease (AD) and related dementias (ADRD; e.g., AD, Lewy body disorders, vascular). Motivated by the Alzheimer's Disease Neuroimaging Initiative's (ADNI) impact, the ADRC Consortium for Clarity in ADRD Research Through Imaging (CLARiTI) was formed. Leveraging existing ADRC infrastructure, CLARiTI will integrate standardized imaging and plasma collection to characterize mixed pathologies and use community‐engaged research methods to ensure that ≥ 25% of the sample is from underrepresented populations (e.g., ethnoculturally minoritized, low education). The resulting ADRD profiles, within a more diverse sample, will provide key resources for ADRCs and an unprecedented, more generalizable publicly available imaging‐plasma dataset. Highlights In vivo detection of mixed pathologies is critical for Alzheimer's disease and related dementias research. The Alzheimer's Disease Research Centers (ADRCs) are uniquely positioned to address gaps related to mixed pathologies. The ADRC Consortium for Clarity in ADRD Research Through Imaging (CLARiTI) will enhance this national program by adding standardized imaging and plasma collection to existing ADRC infrastructure. This effort will provide key resources for ADRCs and an unprecedented publicly available imaging–plasma–neuropath dataset.

Elevated locus coeruleus metabolism provides resilience against cognitive decline in preclinical Alzheimer's disease

Abstract INTRODUCTION Alterations in locus coeruleus’ (LC) metabolic turnover are associated with Alzheimer's disease (AD)‐pathology and cognitive impairment. However, the evolution of these changes across disease stages and their functional relevance remains unknown. METHODS We examined associations of [ 18 F]‐fluorodeoxyglucose positron emission tomography (FDG‐PET) ‐derived LC metabolism with clinical diagnostic status, cerebrospinal fluid (CSF) ‐based AD biomarkers of AD pathology, and cognitive decline in Alzheimer's Disease Neuroimaging Initiative (ADNI) participants ( n  = 604). RESULTS FDG‐PET‐derived LC metabolism was elevated in the earliest preclinical stages and lower in later disease stages. Higher LC metabolism was associated with attenuated memory decline in preclinical stages, particularly in those with low CSF Aβ 42, but not in AD patients with cognitive impairment. DISCUSSION Higher locus coeruleus [ 18 F]‐FDG‐PET‐derived signal in the early preclinical stages of AD can confer cognitive resilience and may reflect increased metabolic activity, whereas later stages are characterized by lower LC FDG‐PET‐derived signal, possibly due to neurodegeneration. Highlights LC FDG‐PET signal is lower in Alzheimer's disease (AD) patients. LC FDG‐PET signal is higher in the preclinical stage of AD. We observed less memory decline in those with higher LC FDG‐PET signal. Higher LC FDG‐PET signal conferred cognitive resilience in preclinical AD.

White matter free water mediates the associations between placental growth factor, white matter hyperintensities, and cognitive status

Abstract INTRODUCTION Placental growth factor (PlGF) may regulate cerebrovascular permeability. We hypothesized that white matter interstitial fluid accumulation, estimated via magnetic resonance imaging (MRI) free water (FW), would explain the associations between elevated PlGF, white matter hyperintensities (WMH), and cognitive impairment. METHODS MarkVCID consortium participants ≥55 years old with plasma PlGF and brain MRI were included. We tested cross‐sectionally whether FW mediated the associations between PlGF and WMH, or PlGF and cognition, measured using the Clinical Dementia Rating (CDR) scale and an executive function (EF) composite (Uniform Data Set version 3 [UDS3]‐EF). RESULTS For 370 participants (mean age 72), a higher PlGF was associated with higher FW, higher WMH, and higher CDR, but not UDS3‐EF. Higher FW was associated with higher WMH, higher CDR, and lower UDS3‐EF. FW explained 26% of the association between PlGF and CDR and 73% of the association between PlGF and WMH. DISCUSSION Elevated PlGF may contribute to WMH and cognitive impairment through white matter FW accumulation. CLINICAL TRIAL REGISTRATION NCT06284213 Highlights PlGF is a promising blood‐based biomarker for vascular cognitive impairment. In MarkVCID, higher PlGF was associated with accumulated white matter FW on MRI. FW mediated the association between higher PlGF and MRI‐visible white matter injury. FW mediated the association between PlGF and worse CDR scale. PlGF may contribute to cognitive dysfunction via accumulated interstitial fluid.

Longitudinal cognitive performance of participants with sporadic early onset Alzheimer's disease from LEADS

Abstract INTRODUCTION Early‐onset Alzheimer's disease (EOAD) manifests prior to the age of 65, and affects 4%–8% of patients with Alzheimer's disease (AD). The current analyses sought to examine longitudinal cognitive trajectories of participants with early‐onset dementia. METHODS Data from 307 cognitively normal (CN) volunteer participants and those with amyloid‐positive EOAD or amyloid‐negative cognitive impairment (EOnonAD) were compared. Cognitive trajectories across a comprehensive cognitive battery spanning 42 months were examined using mixed‐effects modeling. RESULTS The EOAD group displayed worse cognition at baseline relative to EOnonAD and CN groups, and more aggressive declines in cognition over time. The largest effects were observed on measures of executive functioning domains, while memory declines were blunted in EOAD. DISCUSSION EOAD declined 2–4× faster than EOnonAD, and EOAD pathology is not restricted to memory networks. Early identification of deficits is critical to ensure that individuals with sporadic EOAD can be considered for treatment using disease‐modifying medications. Highlights Represents the most comprehensive longitudinal characterization of sporadic EOAD to date. The trajectory of cognitive declines was steep for EOAD participants and worse than for other groups. Executive functioning measures exhibited the greatest declines over time in EOAD.

Developing and applying potentially scalable recruitment strategies to accelerate ADRD research participation of Black adults

Abstract More than 2 million older Americans from underrepresented racial and ethnic minority groups (URGs) have early‐stage Alzheimer's disease and related dementias (ADRD). There are very few scalable recruitment strategies, particularly for Black older adults, to accelerate participation in ADRD research. The Indiana Alzheimer's Disease Research Center (IADRC) and its Community Advisory Board developed and implemented the innovative RAAISE‐D Framework. This Framework informed the creation of community‐first recruitment strategies designed to accelerate participation of Black older adults in ADRD research. Preliminary outcomes from its implementation included the doubling of Black adult enrollment (46, 13.4% to 101, 26.9%) from April 2020 to April 2024. Black adults were more likely to have normal cognition, be female, and ≤ 12 years of education than non‐Hispanic White adults. The RAAISE‐D Framework identified key concepts for URG focused recruitment strategies which successfully accelerated enrollment of Black adults in ADRD research and could be generalized to other URGs. Highlights RAAISE‐D Framework provides adaptable URG recruitment strategies. IADRC CAB‐researcher partnership was the foundation of community‐first methodology. RAAISE‐D Framework doubled the Black enrollment in the IADRC in 4 years.

MRI free water mediates the association between diffusion tensor image analysis along the perivascular space and executive function in four independent middle to aged cohorts

Abstract INTRODUCTION Diffusion tensor image analysis along the perivascular space (DTI‐ALPS) index was proposed for assessing glymphatic clearance function. This study evaluated DTI‐ALPS as a biomarker for cerebral small vessel disease (cSVD) related vascular cognitive impairment and dementia (VCID). METHODS Four independent cohorts were examined. A composite score of executive function (UDS3‐EF) was used to evaluate EF status. The association between the ALPS index and UDS3‐EF scores and the mediator effect of free water in white matter (WM‐FW) on such association was analyzed. RESULTS The ALPS index was significantly associated with UDS3‐EF scores in all cohorts. Additionally, WM‐FW mediates the relationship between the ALPS index and UDS3‐EF scores. DISCUSSION Lower ALPS index may be a surrogate marker of glymphatic dysfunction, which is associated with impaired EF, and this association is mediated by the interstitial fluid (ISF) drainage ISF in WM, providing a clinical rationale for using ALPS index as a biomarker for cSVD‐related VCID. Highlights This is the first study to investigate the mediation role of interstitial FW fraction (WM‐FW) on the relationship between glymphatic clearance (ALPS index) and EF (UDS3‐EF scores) in four independent middle to aged cohorts at risk for cSVD. This study identified that ALPS index was independently associated with UDS3‐EF scores after adjusting for demographics, VRFs, and WM hyperintensity burden and that WM‐FW mediated this association in all middle to aged cohorts. Our findings suggest that in middle to aged individuals, glymphatic dysfunction (reflected by ALPS index) is strongly associated with EF and that this association is mediated by the ISF drainage in WM. This study provides a strong clinical rationale for the use of the ALPS index as a marker of cognitive function in multi‐site observational studies and clinical trials to monitor and prevent VCID.