Kurz gefasst
Fachartikel in der Zeitschrift Himalayan Journal of Ophthalmology, 2024. Verfasst von Anupam Singh. Die Arbeit nennt 16 Literaturangaben.
Abstract (Originalsprache)
Cardiovascular disease (CVD) is a major cause of concern for health care professionals as well as health economists as it is associated with significant morbidity and mortality.[1] CVD is a result of a combination of modifiable risk factors, such as hypertension (HTN), diabetes, dyslipidemia, smoking, and amount of physical activity.[2] Facilitating earlier identification of these risk factors has the potential to significantly alter the negative outcome of CVD. However, medical science is still struggling with universally accepted biomarkers to detect the subclinical stages of CVD.[3] With advances in precision medicine, the focus has shifted to adopting strategies that improve personalized therapy and outcomes.[4] Recently cardio-oculomics has given hope by offering a practical approach to identify and manage cardiovascular risks. It utilizes the retinal microvasculature as a surrogate marker for CVD.[5] Variations in retinal microvasculature in terms of vessel diameter, color, or form can indicate systemic conditions like hypertension or chronic kidney disease.[6,7] Oculomics combines retinal imaging and deep learning algorithms to assess cardiovascular health. An abundance of imaging options like Slit lamp photography, ocular photography, optical coherence tomography (OCT), fundus autofluorescence, retinal fundus photography (RFP), and fluorescein or OCT angiography (OCT A) are all available.[8,9] Because of their affordability, ease of collection, and information depth; RFP and OCT have been widely utilized in oculomics. They offer high-resolution pictures of the optic nerve, retina, and retinal blood vessels. Variations in these domains are strongly linked to the risk of CVD.[10] The integration of the information with deep learning algorithms enhances the accuracy and efficiency of detecting CVD, making the eye a noninvasive, cost-effective, and easily accessible gateway for early cardiovascular health assessment and monitoring. Thus, the eye serves as a clearly defined and information-rich gateway that be accessed noninvasively and is particularly advantageous due to recent advancements in ocular imaging, computational processing, and digital network activity especially focusing on predicting risk factors and discovering biomarkers. Nusinovici et al.[11] created a new deep learning algorithm (DLA) called RetiAge (Singapore Eye research Institute Singapore, Duke-NUS Medical School, Singapore) to measure biological age (BA), which is risk-graded into quartiles for 10-year all-cause mortality, cancer mortality, and CVD mortality independent of chronological age and phenotypic indicators. Compared with patients in the RetiAge first quartile, individuals in the fourth quartile had a 67% increased risk of a 10-year all-cause mortality. They found that the biggest changes occurred in the macula, optic disc edge, and arcade vessels, which explained the expectations. In their internal testing set, RetiAge performed exceptionally well in predicting the likelihood of reaching 65 years or older; but, due to ethnic variances, it did not perform well in their external validation sample.[11] The novelty of establishing BA using less costly and noninvasive retinal photographs and an extension to predict CVD-related mortality and morbidity are the key benefits of this study. The inability to risk stratification based on ethnicity and the inability to process poorer quality RFPs were the major limitations of the study, which may limit the model’s real-world performance. Similarly, Zekavat et al.[12] discovered that fractal branching and microvascular density were associated with an increased risk of death in type 2 diabetes mellitus and HTN cases. Diaz Pinto et al.[13] created a multimodal DLA that was trained on a binary dataset made up of retinal and cardiovascular magnetic resonance images. The algorithm’s excellent accuracy was on par with well-known methods for assessing CVD risk.[13] Zhu et al.[14] demonstrated that DLA based on RFP alone produces estimates comparable to the clinical risk factor-based model in stroke prediction based on targeting decreased central retinal artery equivalent and arteriolar to venule ratio, and that microvasculature may contain signs of ocular end organ damage. This study demonstrated the link between the retinal age difference and stroke risk.[14] Thus, DLAs can therefore be used as indicators for the risk of CVD, the indices produced from the ocular microvasculature may be clinically significant to predict future myocardial infarction. Targeting ocular images with DLA has been more appealing in clinical practice and deployment in recent years because of its noninvasive nature, time and cost savings, and abundance of information.[15] Safety is enhanced by reduced scan radiation, which may boost patient compliance and acceptance for individuals scared of more invasive diagnostic methods. Even with defined techniques and training, inter- and intra-individual grading variability is difficult to eliminate, but DLA performs consistently over similar datasets.[16] Professional graders are expensive to hire, train, and maintain, and there are not enough of them in many parts of the world.[17] Additionally, DLA outperforms humans in terms of availability and repeated work because it is faster and constantly online. By analyzing individual pixels, DLAs allow for the detection of important details that may go unnoticed by the naked eye in the field of research and development. Examples include elevated stiffness in retinal microvasculature and alterations in venules, which are linked to increased buildup of coronary artery calcium.[18] These recently discovered biomarkers and correlations have the potential to contribute to the development of innovative theories.[12] The confounders, such as age-related macular degeneration, have been demonstrated to reduce the predictive capacity of DLA. This is likely due to the obliteration of the microvasculature signals that DLAs rely on.[13] The lack of full-stack software and a suitably large heterogeneous data collection for training can further contribute to the challenge. With the current advancements in ocular imaging, computational processing, and digital network traffic, the eye is a well-defined, information-rich, and noninvasive portal for insights into CVD. Appropriately, patient requirements and clinical medicine are the areas where DLAs are expected to be most useful due to the growing number of elements taken into account in clinical practice. This is particularly true when combining DLAs’ analytical capabilities with information-rich imaging modalities to reframe what the eye can disclose about CVD. The cardio-oculomics is underexplored yet has enormous promise for early disease detection, prediction, and screening. Thus, in the future interdisciplinary DLA development and optimization will improve patient outcomes and medical resource allocations. DLA for cardiac-oculomics risk profiling is fascinating, bothersome, and potentially impactful, despite its unresolved issues. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.
Zitieren
APA
Singh, A. (2024). Cardio-oculomics. Himalayan Journal of Ophthalmology. https://doi.org/10.4103/hjo.hjo_14_24
Vancouver
Singh A. Cardio-oculomics. Himalayan Journal of Ophthalmology. 2024. doi:10.4103/hjo.hjo_14_24
BibTeX
@article{Singh2024,
author = {Singh, Anupam},
title = {Cardio-oculomics},
journal = {Himalayan Journal of Ophthalmology},
year = {2024},
doi = {10.4103/hjo.hjo_14_24}
}
EUVIMED