Deep learning predicts prevalent and incident Parkinson's disease from UK Biobank fundus imaging.
Ontology highlight
ABSTRACT:
SUBMITTER: Tran C
PROVIDER: S-EPMC10864361 | biostudies-literature | 2024 Feb
REPOSITORIES: biostudies-literature
Publications
Deep learning predicts prevalent and incident Parkinson's disease from UK Biobank fundus imaging.
Tran Charlie C Shen Kai K Liu Kang K Ashok Akshay A Ramirez-Zamora Adolfo A Chen Jinghua J Li Yulin Y Fang Ruogu R
Scientific reports 20240213 1
Parkinson's disease is the world's fastest-growing neurological disorder. Research to elucidate the mechanisms of Parkinson's disease and automate diagnostics would greatly improve the treatment of patients with Parkinson's disease. Current diagnostic methods are expensive and have limited availability. Considering the insidious and preclinical onset and progression of the disease, a desirable screening should be diagnostically accurate even before the onset of symptoms to allow medical interven ...[more]
Similar Datasets
Project description:ObjectivesUK Biobank is a UK-wide cohort of 502,655 people aged 40-69, recruited from National Health Service registrants between 2006-10, with healthcare data linkage. Type 2 diabetes is a key exposure and outcome. We developed algorithms to define prevalent and incident diabetes for UK Biobank. The algorithms will be implemented by UK Biobank and their results made available to researchers on request.MethodsWe used UK Biobank self-reported medical history and medication to assign prevalent diabetes and type, and tested this against linked primary and secondary care data in Welsh UK Biobank participants. Additionally, we derived and tested algorithms for incident diabetes using linked primary and secondary care data in the English Clinical Practice Research Datalink, and ran these on secondary care data in UK Biobank.Results and significanceFor prevalent diabetes, 0.001% and 0.002% of people classified as "diabetes unlikely" in UK Biobank had evidence of diabetes in their primary or secondary care record respectively. Of those classified as "probable" type 2 diabetes, 75% and 96% had specific type 2 diabetes codes in their primary and secondary care records. For incidence, 95% of people with the type 2 diabetes-specific C10F Read code in primary care had corroborative evidence of diabetes from medications, blood testing or diabetes specific process of care codes. Only 41% of people identified with type 2 diabetes in primary care had secondary care evidence of type 2 diabetes. In contrast, of incident cases using ICD-10 type 2 diabetes specific codes in secondary care, 77% had corroborative evidence of diabetes in primary care. We suggest our definition of prevalent diabetes from UK Biobank baseline data has external validity, and recommend that specific primary care Read codes should be used for incident diabetes to ensure precision. Secondary care data should be used for incident diabetes with caution, as around half of all cases are missed, and a quarter have no corroborative evidence of diabetes in primary care.
| S-EPMC5025160 | biostudies-literature
Project description:BackgroundCardiovascular disease is a leading cause of global death. Prospective population-based studies have found that changes in retinal microvasculature are associated with the development of coronary artery disease. Recently, artificial intelligence deep learning (DL) algorithms have been developed for the fully automated assessment of retinal vessel calibres.MethodsIn this study, we validate the association between retinal vessel calibres measured by a DL system (Singapore I Vessel Assessment) and incident myocardial infarction (MI) and assess its incremental performance in discriminating patients with and without MI when added to risk prediction models, using a large UK Biobank cohort.ResultsRetinal arteriolar narrowing was significantly associated with incident MI in both the age, gender and fellow calibre-adjusted (HR=1.67 (95% CI: 1.19 to 2.36)) and multivariable models (HR=1.64 (95% CI: 1.16 to 2.32)) adjusted for age, gender and other cardiovascular risk factors such as blood pressure, diabetes mellitus (DM) and cholesterol status. The area under the receiver operating characteristic curve increased from 0.738 to 0.745 (p=0.018) in the age-gender-adjusted model and from 0.782 to 0.787 (p=0.010) in the multivariable model. The continuous net reclassification improvements (NRIs) were significant in the age and gender-adjusted (NRI=21.56 (95% CI: 3.33 to 33.42)) and the multivariable models (NRI=18.35 (95% CI: 6.27 to 32.61)). In the subgroup analysis, similar associations between retinal arteriolar narrowing and incident MI were observed, particularly for men (HR=1.62 (95% CI: 1.07 to 2.46)), non-smokers (HR=1.65 (95% CI: 1.13 to 2.42)), patients without DM (HR=1.73 (95% CI: 1.19 to 2.51)) and hypertensive patients (HR=1.95 (95% CI: 1.30 to 2.93)) in the multivariable models.ConclusionOur results support DL-based retinal vessel measurements as markers of incident MI in a predominantly Caucasian population.
| S-EPMC10961540 | biostudies-literature
Project description:To examine the association and modifiable risk factors between grip strength (GS) and Parkinson's disease (PD) incidence considering genetic factors, a total of 411,648 individuals without PD at baseline from the UK Biobank were included. GS was measured by a hydraulic dynamometer. The polygenic risk score assessed the genetic predisposition. Multivariable Cox regression models were performed. During a median follow-up of 12.3 years, 2409 individuals developed PD. Compared with those with high GS, low-GS individuals had a 58.5% increased risk of PD (42.7%-76.1%), and 16.3% of this excess risk could be explained by adjusted risk factors. Low GS and high genetic predisposition contribute to the highest PD risk in an additive interaction. We observed that low GS was associated with higher PD incidence, particularly among individuals with high genetic predisposition. In addition to enhancing GS, interventions targeting risk factors (e.g., unhealthy lifestyles) might also reduce the excess risk.
| S-EPMC11494040 | biostudies-literature
Project description:BackgroundAdverse muscle composition (MC) as measured by magnetic resonance imaging has previously been linked to poor function, comorbidity, and increased hospitalization. The aim of this study was to investigate if adverse MC predicts all-cause mortality using data from UK Biobank.MethodsThere were 40 178 participants scanned using a 6 min magnetic resonance imaging protocol. Images were analysed for thigh fat-tissue free muscle volume and muscle fat infiltration (MFI) using AMRA® Researcher (AMRA Medical, Linköping, Sweden). For each participant, a sex, weight, and height invariant muscle volume z-score was calculated. Participants were partitioned into four MC groups: (i) normal MC, (ii) only low muscle volume [<25th percentile for muscle volume z-score (population wide)], (iii) only high MFI [>75th percentile (population wide, sex-specific)], and (iv) adverse MC (low muscle volume z-score and high MFI). Association of MC groups with mortality was investigated using Cox proportional-hazard modelling with normal MC as referent (unadjusted and adjusted for low hand grip strength, sex, age, body mass index, previous diagnosis of disease (cancer, type 2 diabetes and coronary heart disease), lifestyle, and socioeconomic factors (smoking, alcohol consumption, physical activity, and Townsend deprivation index).ResultsMuscle composition measurements were complete for 39 804 participants [52% female, mean (SD) age 64.2 (7.6) years and body mass index 26.4 (4.4) kg/m2 ]. Three hundred twenty-eight deaths were recorded during a follow-up period of 2.9 (1.4) years after imaging. At imaging, adverse MC was detected in 10.5% of participants. The risk of death from any cause in adverse MC compared with normal MC was 3.71 (95% confidence interval 2.81-4.91, P < 0.001). Only low muscle volume and only high MFI were independently associated with all-cause mortality [1.58 (1.13-2.21), P = 0.007, and 2.02 (1.51-2.71), P < 0.001, respectively]. Adjustment of low hand grip strength [1.77 (1.28-2.44), P < 0.001] did not attenuate the associations with any of the MC groups. In the fully adjusted model, adverse MC and only high MFI remained significant (P < 0.001 and P = 0.020) while the association with only low muscle volume was attenuated to non-significance (P = 0.560). The predictive performance of adverse MC [1.96 (1.42-2.71), P < 0.001] was comparable with that of previous cancer diagnosis [1.93 (1.47-2.53), P < 0.001] and smoking [1.71 (1.02-2.84), P = 0.040]. Low hand grip strength was borderline non-significant [1.34 (0.96-1.88), P = 0.090].ConclusionsAdverse MC was a strong and independent predictor of all-cause mortality. Sarcopenia guidelines can be strengthened by including cut-offs for myosteatosis enabling detection of adverse MC.
| S-EPMC8718078 | biostudies-literature
Project description:ObjectiveEvidence on the individual and combined relationship of physical activity (PA) and fish oil supplement use on the incidence of Parkinson's disease (PD) risk remains lacking.Materials and methodsThis UK population-based prospective cohort study, involving 385,275 UK Biobank participants, collected PA and fish oil supplement data via touchscreen questionnaires. Using Cox proportional hazards models and restricted cubic splines to examined the associations between use of fish oil supplements, PA and PD risk.ResultsDuring a median 12.52-year follow-up, 2,131 participants incident PD. Analysis showed that fish oil supplement users had a lower PD risk [hazard ratio (HR), 0.89; 95% confidence interval (CI), 0.82-0.98]. The adjusted HRs for the PD incidence were 0.96 (95% CI, 0.95-0.98) for total PA; 0.93 (95% CI, 0.90-0.96) for moderate PA; 0.95 (95% CI, 0.91-0.99) for vigorous PA and 0.93 (95% CI, 0.89-0.98) for walking activity. Significant interactions were found between fish oil supplement use and total PA (P for interaction = 0.011), moderate PA (P for interaction = 0.015), and walking activity (P for interaction = 0.029) in relation to PD incidence.ConclusionBoth fish oil supplement use and PA were associated with a reduced risk of PD, and the effect of PA in reducing the risk of PD was more pronounced when fish oil supplement was used.
| S-EPMC10859434 | biostudies-literature
Project description:BackgroundNutritional status influences immunity but its specific association with susceptibility to COVID-19 remains unclear. We examined the association of specific dietary data and incident COVID-19 in the UK Biobank (UKB).MethodsWe considered UKB participants in England with self-reported baseline (2006-2010) data and linked them to Public Health England COVID-19 test results-performed on samples from combined nose/throat swabs, using real time polymerase chain reaction (RT-PCR)-between March and November 2020. Baseline diet factors included breastfed as baby and specific consumption of coffee, tea, oily fish, processed meat, red meat, fruit, and vegetables. Individual COVID-19 exposure was estimated using the UK's average monthly positive case rate per specific geo-populations. Logistic regression estimated the odds of COVID-19 positivity by diet status adjusting for baseline socio-demographic factors, medical history, and other lifestyle factors. Another model was further adjusted for COVID-19 exposure.ResultsEligible UKB participants (n = 37,988) were 40 to 70 years of age at baseline; 17% tested positive for COVID-19 by SAR-CoV-2 PCR. After multivariable adjustment, the odds (95% CI) of COVID-19 positivity was 0.90 (0.83, 0.96) when consuming 2-3 cups of coffee/day (vs. <1 cup/day), 0.88 (0.80, 0.98) when consuming vegetables in the third quartile of servings/day (vs. lowest quartile), 1.14 (1.01, 1.29) when consuming fourth quartile servings of processed meats (vs. lowest quartile), and 0.91 (0.85, 0.98) when having been breastfed (vs. not breastfed). Associations were attenuated when further adjusted for COVID-19 exposure, but patterns of associations remained.ConclusionsIn the UK Biobank, consumption of coffee, vegetables, and being breastfed as a baby were favorably associated with incident COVID-19; intake of processed meat was adversely associated. Although these findings warrant independent confirmation, adherence to certain dietary behaviors may be an additional tool to existing COVID-19 protection guidelines to limit the spread of this virus.
| S-EPMC8234071 | biostudies-literature
Project description:ImportanceRetinal diseases are the leading cause of irreversible blindness worldwide, and timely detection contributes to prevention of permanent vision loss, especially for patients in rural areas with limited medical resources. Deep learning systems (DLSs) based on fundus images with a 45° field of view have been extensively applied in population screening, while the feasibility of using ultra-widefield (UWF) fundus image-based DLSs to detect retinal lesions in patients in rural areas warrants exploration.ObjectiveTo explore the performance of a DLS for multiple retinal lesion screening using UWF fundus images from patients in rural areas.Design, setting, and participantsIn this diagnostic study, a previously developed DLS based on UWF fundus images was used to screen for 5 retinal lesions (retinal exudates or drusen, glaucomatous optic neuropathy, retinal hemorrhage, lattice degeneration or retinal breaks, and retinal detachment) in 24 villages of Yangxi County, China, between November 17, 2020, and March 30, 2021.InterventionsThe captured images were analyzed by the DLS and ophthalmologists.Main outcomes and measuresThe performance of the DLS in rural screening was compared with that of the internal validation in the previous model development stage. The image quality, lesion proportion, and complexity of lesion composition were compared between the model development stage and the rural screening stage.ResultsA total of 6222 eyes in 3149 participants (1685 women [53.5%]; mean [SD] age, 70.9 [9.1] years) were screened. The DLS achieved a mean (SD) area under the receiver operating characteristic curve (AUC) of 0.918 (0.021) (95% CI, 0.892-0.944) for detecting 5 retinal lesions in the entire data set when applied for patients in rural areas, which was lower than that reported at the model development stage (AUC, 0.998 [0.002] [95% CI, 0.995-1.000]; P < .001). Compared with the fundus images in the model development stage, the fundus images in this rural screening study had an increased frequency of poor quality (13.8% [860 of 6222] vs 0%), increased variation in lesion proportions (0.1% [6 of 6222]-36.5% [2271 of 6222] vs 14.0% [2793 of 19 891]-21.3% [3433 of 16 138]), and an increased complexity of lesion composition.Conclusions and relevanceThis diagnostic study suggests that the DLS exhibited excellent performance using UWF fundus images as a screening tool for 5 retinal lesions in patients in a rural setting. However, poor image quality, diverse lesion proportions, and a complex set of lesions may have reduced the performance of the DLS; these factors in targeted screening scenarios should be taken into consideration in the model development stage to ensure good performance.
| S-EPMC10587822 | biostudies-literature
Project description:Background: Optical coherence tomography is a powerful modality to assess atherosclerotic lesions, but detecting lesions in high-resolution OCT is challenging and requires expert knowledge. Deep-learning algorithms can be used to automatically identify atherosclerotic lesions, facilitating identification of patients at risk. We trained a deep-learning algorithm (DeepAD) with co-registered, annotated histopathology to predict atherosclerotic lesions in optical coherence tomography (OCT). Methods: Two datasets were used for training DeepAD: (i) a histopathology data set from 7 autopsy cases with 62 OCT frames and co-registered histopathology for high quality manual annotation and (ii) a clinical data set from 51 patients with 222 OCT frames in which manual annotations were based on clinical expertise only. A U-net based deep convolutional neural network (CNN) ensemble was employed as an atherosclerotic lesion prediction algorithm. Results were analyzed using intersection over union (IOU) for segmentation. Results: DeepAD showed good performance regarding the prediction of atherosclerotic lesions, with a median IOU of 0.68 ± 0.18 for segmentation of atherosclerotic lesions. Detection of calcified lesions yielded an IOU = 0.34. When training the algorithm without histopathology-based annotations, a performance drop of >0.25 IOU was observed. The practical application of DeepAD was evaluated retrospectively in a clinical cohort (n = 11 cases), showing high sensitivity as well as specificity and similar performance when compared to manual expert analysis. Conclusion: Automated detection of atherosclerotic lesions in OCT is improved using a histopathology-based deep-learning algorithm, allowing accurate detection in the clinical setting. An automated decision-support tool based on DeepAD could help in risk prediction and guide interventional treatment decisions.
| S-EPMC8713728 | biostudies-literature
Project description:BackgroundFundus Autofluorescence (FAF) is a valuable imaging technique used to assess metabolic alterations in the retinal pigment epithelium (RPE) associated with various age-related and disease-related changes. The practical uses of FAF are ever-growing. This study aimed to evaluate the effectiveness of a generative deep learning (DL) model in translating color fundus (CF) images into synthetic FAF images and explore its potential for enhancing screening of age-related macular degeneration (AMD).MethodsA generative adversarial network (GAN) model was trained on pairs of CF and FAF images to generate synthetic FAF images. The quality of synthesized FAF images was assessed objectively by common generation metrics. Additionally, the clinical effectiveness of the generated FAF images in AMD classification was evaluated by measuring the area under the curve (AUC), using the LabelMe dataset.ResultsA total of 8410 FAF images from 2586 patients were analyzed. The synthesized FAF images exhibited an impressive objectively assessed quality, achieving a multi-scale structural similarity index (MS-SSIM) of 0.67. When evaluated on the LabelMe dataset, the combination of generated FAF images and CF images resulted in a noteworthy improvement in AMD classification accuracy, with the AUC increasing from 0.931 to 0.968.ConclusionsThis study presents the first attempt to use a generative deep learning model to create authentic and high-quality FAF images from CF images. The incorporation of the translated FAF images on top of CF images improved the accuracy of AMD classification. Overall, this study presents a promising approach to enhance large-scale AMD screening.
| S-EPMC10696390 | biostudies-literature
Project description:Dealing with confounds is an essential step in large cohort studies to address problems such as unexplained variance and spurious correlations. UK Biobank is a powerful resource for studying associations between imaging and non-imaging measures such as lifestyle factors and health outcomes, in part because of the large subject numbers. However, the resulting high statistical power also raises the sensitivity to confound effects, which therefore have to be carefully considered. In this work we describe a set of possible confounds (including non-linear effects and interactions that researchers may wish to consider for their studies using such data). We include descriptions of how we can estimate the confounds, and study the extent to which each of these confounds affects the data, and the spurious correlations that may arise if they are not controlled. Finally, we discuss several issues that future studies should consider when dealing with confounds.
| S-EPMC7610719 | biostudies-literature