A literature-derived dataset on risk factors for dry eye disease

Abstract

Dry eye disease (DED) is a common disease associated with disorder of tear secretion. Research on risk factors for DED, such as depression, arthritis, thyroid disease, stroke and diabetes, is important to facilitate its diagnosis and prognosis. We created a dataset on risk factors for DED (DrDED) with public access that can provide up-to-date and validated data acquired from systematically searched and screened, high-quality studies. The established DrDED contained 119 studies published between 2000 and 2022. The range of the study sample size was from 43 to 4,871,504. The study types were, as follows: cross-sectional (n = 92), retrospective cohort (n = 9), prospective cohort (n = 10), and case-control (n = 8) studies. Data from eligible studies were collected and presented for the present study, including the publication information, study characteristics, definition and prevalence of the disease, and risk factors for DED, together with the strength of association. With the publication of new relevant studies, the DrDED will be updated, and the data will be made accessible to the users. Design Type(s) Dataset creation objective Measurement Type(s) Patient outcome • scientific publication • risk factors • dry eye disease Technology Type(s) Digital curation • documenting • meta-analysis Factor Type(s) Depression • arthritis • thyroid disease • stroke disease • diabetes Sample Characteristic(s) Homo sapiens • dry eye disease • global

MonoNet: enhancing interpretability in neural networks via monotonic features

“…Since these datasets are relatively simple, we do not compare MonoNet to other unconstrained deep models, but limit the comparison to two interpretable models, ( Ustun and Rudin, 2017 ) and decision trees ( Quinlan, 1986 ). In the vision domain, we test our model on the biomedical image datasets provided by Yang et al (2021 , 2023) . Since the more traditional machine learning models are not directly scalable to this domain, we compare MonoNet to established unconstrained neural network models [i.e.…”

“…ResNets ( He et al , 2016 ), as well as automatically optimized models as provided by auto-sklearn ( Feurer et al , 2015 ), AutoKeras ( Jin et al , 2019 ) and Google AutoML Vision. All results, apart from MonoNets, are taken from Yang et al (2023) . Best performing model on each dataset are indicated with bold face.…”

“…Equation (4)] , we can infer the directionality and strength of the monotonic relationship. As a concrete example, we analyze a Monotonic CNN trained on PathMNIST ( Yang et al , 2023 ), a down-sampled version of the histological dataset provided by Kather et al (2019) to study colorectal cancer. Figure 7 shows 1 of the 256 convolutional filters (right) and its location on the sample we are interpreting (left).…”

Measuring Bedload Motion Time at Second Resolution Using Benford's Law on Acoustic Data

“…All data and MATLAB code analyzed in this study are available at Yang (2023). The spectrograms for the sounds of bedload transport and background noise, including the spectrograms for bedload transport events defined in 2019 and 2021 are available at Yang (2024).…”

Assessment of leukocyte and systemic inflammation index ratios in dyslipidemia patients with dry eye disease: a retrospective case‒control study

“…DLP is a significant contributor to coronary artery disease [ 3 ]. In addition, DLP has been associated with chronic inflammatory diseases such as diabetes [ 5 ], thyroid disease [ 6 ], and dry eye disease (DED) [ 7 ]. DLP has a high prevalence worldwide; for instance, 19.7% of Korean adolescents, 27% of American adolescents, and 34% of Chinese adults have been reported to have DLP [ 8 ].…”