Investigation of the Retinal Biomarkers of Alzheimer’s Disease and Atherosclerosis Using Hyperspectral Images
Bibliographic record
Abstract
Le fait que l'oeil puisse etre visualise de maniere non invasive ouvre des possibilites de mesure de biomarqueurs pour le diagnostic de conditions a long terme. Selon de nombreuses etudes, plusieurs maladies cardiovasculaires et neurodegeneratives telles que la maladie d’Alzheimer (AD) et l’atherosclerose (ATH) se manifestent dans la retine sous forme de modifications morphologiques pathologiques et / ou vasculaires. Des methodes d'imagerie oculaire en deux dimensions et des techniques de tomographie par coherence optique (OCT) en trois dimensions ont ete developpees pour fournir des descriptions des structures retiniennes. Cependant, les images acquises par ces techniques permettent principalement de mesurer les caracteristiques spatiales et pas la variance relative de l’intensite des pixels sur differentes longueurs d’onde, de sorte que d’importantes caracteristiques liees aux tissus peuvent encore rester a decouvrir. Dans cette etude, une camera retinienne metabolique hyperspectrale (MHRC) a ete utilisee pour permettre l'acquisition d'une serie d'images retiniennes obtenues a des longueurs d'onde specifiques couvrant le spectre du visible au proche infrarouge (NIR). Dans cette technique, le facteur de transmission, l'absorption et la diffusion de la lumiere sont refletes dans le spectre de la lumiere emise par le tissu. Par consequent, non seulement les caracteristiques spatiales communes mais egalement les « signatures spectrales » de biomolecules pourraient etre revelees. Cela aide a trouver une plus grande variete de caracteristiques spatiales / spectrales pour une investigation plus precise des biomarqueurs retiniens des maladies. En ce qui concerne les couts et les limites associes aux diagnostics actuels de l’AD et de l’ATH, le but de cette these etait d’analyser le contenu en informations d’images retiniennes hyperspectrales riches en donnees dans le but de caracteriser des informations discriminantes cachees liees aux tissus afin d’identifier des biomarqueurs possibles de ces deux maladies. A cette fin, une combinaison de caracteristiques vasculaires et de mesures de textures spatiales-spectrales ont ete extraites de differentes regions anatomiques de la retine. Dans le contexte de la maladie d'Alzheimer, des images retiniennes de 20 cas presentant une alteration cognitive et de 26 cas normaux cognitivement ont ete acquises a l'aide de la camera MHRC. Le statut amyloide cerebral a ete determine a partir de lectures binaires effectuees par un panel de 3 experts noteurs ayant participe a des etudes de TEP au 18F-Florbetaben. Des caracteristiques de l’image retinienne ont ete calculees, notamment la tortuosite et le diametre des vaisseaux, ainsi que les mesures de textures spatiales-spectrales sur les arterioles, les veinules et le tissu environnant. Les veinules retiniennes des sujets amyloides positifs (Aβ +) ont presente une tortuosite moyenne plus elevee par rapport aux sujets amyloides negatifs (Aβ-). Le diametre arteriolaire des sujets Aβ + s'est avere superieur a celui des sujets Aβ- dans une zone adjacente a la tete du nerf optique. De plus, une difference significative entre les mesures de texture construites sur les arterioles retiniennes et leurs regions adjacentes a ete observee chez les sujets Aβ + par rapport aux Aβ-. Dans le contexte de l'ATH, 60 images retiniennes de 30 ATH probables sur le plan clinique et 30 cas de controle ont ete acquises. Les criteres d'inclusion pour les sujets souffrant d'ATH comprenaient: l'infarctus du myocarde; angiographie coronaire montrant au moins une stenose coronaire (plus de 50%); et / ou une angioplastie coronaire; et /ou pontage coronaire. Les arterioles retiniennes des sujets ATH ont montre un retrecissement significatif par rapport aux sujets temoins. En outre, une difference significative entre les mesures de textures d'images prises sur les arterioles et les veinules retiniennes et leurs regions adjacentes a ete trouvee entre les sujets ATH et les sujets temoins. Nos etudes transversales ont montre que l’analyse hyperspectrale des images retiniennes pouvait discerner avec une precision acceptable l’AD et l’ATH des sujets temoins correspondants.----------ABSTRACT The fact that eye can be visualized non-invasively, opens up possibilities to measure biomarkers for diagnosis of long-term conditions. A significant body of literature has demonstrated that many of the neurodegenerative and cardiovascular diseases such as Alzheimer’s disease (AD) and atherosclerosis (ATH) manifest themselves in retina as pathological and/or vasculature morphological changes. Methods for two-dimensional fundus imaging and techniques for three-dimensional optical coherence tomography (OCT) have been developed to provide descriptions of retinal structures. However, images acquired by these techniques mostly allow for measuring the spatial characteristics of the tissue and lack of the relative variances across differing wavelengths, thus important spectral features may remain uncovered. In this study, a Metabolic Hyperspectral Retinal Camera (MHRC) was used that permits the acquisition of a series of retinal images obtained at specific wavelengths covering the visible and near infrared (NIR) spectrum. In this technique, light transmittance, absorption, and scatter are reflected in the spectrum of light emitted from the tissue. Use of MHRC in this study was aimed to extract not only the common spatial features but also “spectral signatures” of biomolecules in retinal tissue. Regarding the costs and limitations of the current diagnostic methods for AD and ATH, the purpose of this thesis was to analyze the information content of data-rich hyperspectral retinal images to characterize tissue-related discriminatory information to identify possible biomarkers of Alzheimer’s disease and atherosclerosis. To this end, a combination of vascular features and spatial/spectral texture measures were extracted from different anatomical regions of the retina. In the context of AD, retinal images from 20 cognitively impaired and 26 cognitively unimpaired cases were acquired using MHRC. The cerebral amyloid status was determined from binary reads by a panel of three expert raters on 18F-Florbetaben PET studies. Our approach did not aim to visualize directly Aβ deposits in the retina but rather to determine a likely amyloid status based on sets of retinal image features highly correlated with the cerebral amyloid status. Retinal image features were calculated including vessels’ tortuosity and diameter. Spatial/spectral texture measures over arterioles, venules, and tissue around were also extracted. Retinal venules of amyloid positive subjects (Aβ+) showed a higher mean tortuosity compared to the amyloid negative (Aβ-) subjects. Arteriolar diameter of Aβ+ subjects was found to be higher than the Aβ- subjects in a zone adjacent to the optical nerve head. Furthermore, a significant difference between spatial/spectral texture measures built over retinal arterioles and surrounding tissues were observed in Aβ+ subjects when compared to the Aβ-. In the context of ATH, 60 retinal images from 30 clinically probable ATH and 30 control cases were acquired. Inclusion criteria for subjects suffering from ATH included: myocardial infarction; coronary angiography showing at least one coronary stenosis (more than 50%); and/or coronary angioplasty; and/or coronary bypass. Retinal arterioles of ATH subjects showed a significant narrowing when compared to control subjects. Moreover, a significant difference between image texture measures taken over retinal arterioles and retinal venules and their adjacent regions was observed between ATH subjects and control subjects. Our cross-sectional studies have shown that hyperspectral retinal image analysis could be used to discriminate AD and ATH from corresponding control subjects based on a non-invasive eye scan.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".