Comparative evaluation of interpretation methods in surface-based age prediction for neonates.

Significant changes in brain morphology occur during the third trimester of gestation. The capability of deep learning in leveraging these morphological features has enhanced the accuracy of brain age predictions for this critical period. Yet, the opaque nature of deep learning techniques, often described as "black box" approaches, limits their interpretability, posing challenges in clinical applications. Traditional interpretable methods developed for computer vision and natural language processing may not directly translate to the distinct demands of neuroimaging. In response, our research evaluates the effectiveness and adaptability of two interpretative methods-regional age prediction and the perturbation-based saliency map approach-for predicting the brain age of neonates. Analyzing 664 T1 MRI scans with the NEOCIVET pipeline to extract brain surface and cortical features, we assess how these methods illuminate key brain regions for age prediction, focusing on technical analysis with clinical insight. Through a comparative analysis of the saliency index (SI) with relative brain age (RBA) and the examination of structural covariance networks, we uncover the saliency index's enhanced ability to pinpoint regions vital for accurate indication of clinical factors. Our results highlight the advantages of perturbation techniques in addressing the complexities of medical data, steering clinical interventions for premature neonates towards more personalized and interpretable approaches. This study not only reveals the promise of these methods in complex medical scenarios but also offers a blueprint for implementing more interpretable and clinically relevant deep learning models in healthcare settings.

A Turbidity-Compensation Method for Nitrate Measurement Based on Ultraviolet Difference Spectroscopy.

To solve the problem that turbidity in water has a significant effect on the spectra of nitrate and reduces the accuracy of nitrate detection, a turbidity-compensation method for nitrate measurement based on ultraviolet difference spectra is proposed. The effect of turbidity on the absorption spectra of nitrate was studied by using the difference spectra of the mixed solution and a nitrate solution. The results showed that the same turbidity had different effects on the absorbance of different concentrations of nitrate. The change in absorbance due to turbidity decreased with an increase in the nitrate concentration at wavelengths from 200 nm to 230 nm, although this change was constant when the wavelength was greater than 230 nm. On the basis of this characteristic, we combined the residual sum of squares (RSS) and interval partial least squares (iPLS) to select wavelengths of 230-240 nm as the optimal modeling interval. Furthermore, the turbidity-compensation model was established by the linear fitting of the difference spectra of various levels of turbidity. The absorption spectra of the nitrate were extracted by subtracting the turbidity-compensation curve from the original spectra of the water samples, and the nitrate concentration was calculated by using a partial least squares (PLS)-based nitrate-prediction model. The experimental results showed that the average relative error of the nitrate predictions was reduced by 50.33% to 1.33% by the proposed turbidity-compensation method. This indicated that this method can better correct the deviation in nitrate's absorbance caused by turbidity and improve the accuracy of nitrate predictions.

Relationships between accelerometer-measured and multiple sclerosis: a 2-sample Mendelian randomization study.

BACKGROUND: Observational studies suggest that physical activity (PA) can independently modify the risk of developing multiple sclerosis (MS). OBJECTIVE: To investigate the causal effect of PA on MS by Mendelian randomization (MR) approaches. METHODS: Through a genome-wide association study including 91,105 participants from UK Biobank, we obtained 5 single-nucleotide polymorphisms (SNPs) associated with accelerometer-measured PA (P < 5 × 10-8). Summary-level data for MS were obtained from a meta-analysis, incorporating 14,802 subjects with MS and 26,703 healthy controls of European ancestry. MR analyses were performed using the inverse-variance-weighted method, weighted median estimator, and MR-PRESSO method. Additional analyses were further performed using MR-Egger intercept and Cochran's Q statistic to verify the robustness of our findings. RESULTS: We failed to detect a causal effect of PA on MS (OR, 0.60; 95% confidence interval [CI], 0.30-1.20; P = 0.15) per in the random-effects IVW analysis. Additional MR methods yielded consistent results. MR-Egger regression suggested no evidence of horizontal pleiotropy (Intercept = 0.14, P = 0.21) and there seemed no substantial heterogeneity (I2 = 29.8%, P = 0.22) among individual SNPs. CONCLUSION: Our findings suggest that enhancing PA might not modify the risk of developing MS independent of established risk factors.

Multiple marine algae identification based on three-dimensional fluorescence spectroscopy and multi-label convolutional neural network.

Accurate identification of algal populations plays a pivotal role in monitoring seawater quality. Fluorescence-based techniques are effective tools for quickly identifying different algae. However, multiple coexisting algae and their similar photosynthetic pigments can constrain the efficacy of fluorescence methods. This study introduces a multi-label classification model that combines a specific Excitation-Emission matric convolutional neural network (EEM-CNN) with three-dimensional (3D) fluorescence spectroscopy to detect single and mixed algal samples. Spectral data can be input directly into the model without transforming into images. Rectangular convolutional kernels and double convolutional layers are applied to enhance the extraction of balanced and comprehensive spectral features for accurate classification. A dataset comprising 3D fluorescence spectra from eight distinct algae species representing six different algal classes was obtained, preprocessed, and augmented to create input data for the classification model. The classification model was trained and validated using 4448 sets of test samples and 60 sets of test samples, resulting in an accuracy of 0.883 and an F1 score of 0.925. This model exhibited the highest recognition accuracy in both single and mixed algae samples, outperforming comparative methods such as ML-kNN and N-PLS-DA. Furthermore, the classification results were extended to three different algae species and mixed samples of skeletonema costatum to assess the impact of spectral similarity on multi-label classification performance. The developed classification models demonstrated robust performance across samples with varying concentrations and growth stages, highlighting CNN's potential as a promising tool for the precise identification of marine algae.

Effective correction of dissolved organic carbon interference in nitrate detection using ultraviolet spectroscopy combined with the equivalent concentration offset method.

Nitrate contamination in water sources poses a substantial environmental and health risk. However, accurate detection of nitrate in water, particularly in the presence of dissolved organic carbon (DOC) interference, remains a significant analytical challenge. This study investigates a novel approach for the reliable detection of nitrate in water samples with varying levels of DOC interference based on the equivalent concentration offset method. The characteristic wavelengths of DOC were determined based on the first-order derivatives, and a nitrate concentration prediction model based on partial least squares (PLS) was established using the absorption spectra of nitrate solutions. Subsequently, the absorption spectra of the nitrate solutions were subtracted from that of the nitrate-DOC mixed solutions to obtain the difference spectra. These difference spectra were introduced into the nitrate prediction model to calculate the equivalent concentration offset values caused by DOC. Finally, a DOC interference correction model was established based on a binary linear regression between the absorbances at the DOC characteristic wavelengths and the DOC-induced equivalent concentration offset values of nitrate. Additionally, a modeling wavelength selection algorithm based on a sliding window was proposed to ensure the accuracy of the nitrate concentration prediction model and the equivalent concentration offset model. The experimental results demonstrated that by correcting the DOC-induced offsets, the relative error of nitrate prediction was reduced from 94.44% to 3.36%, and the root mean square error of prediction was reduced from 1.6108 mg L-1 to 0.1037 mg L-1, which is a significant correction effect. The proposed method applied to predict nitrate concentrations in samples from two different water sources shows a certain degree of comparability with the standard method. It proves that this method can effectively correct the deviations in nitrate measurements caused by DOC and improve the accuracy of nitrate measurement.

A multicenter study on two-stage transfer learning model for duct-dependent CHDs screening in fetal echocardiography.

Duct-dependent congenital heart diseases (CHDs) are a serious form of CHD with a low detection rate, especially in underdeveloped countries and areas. Although existing studies have developed models for fetal heart structure identification, there is a lack of comprehensive evaluation of the long axis of the aorta. In this study, a total of 6698 images and 48 videos are collected to develop and test a two-stage deep transfer learning model named DDCHD-DenseNet for screening critical duct-dependent CHDs. The model achieves a sensitivity of 0.973, 0.843, 0.769, and 0.759, and a specificity of 0.985, 0.967, 0.956, and 0.759, respectively, on the four multicenter test sets. It is expected to be employed as a potential automatic screening tool for hierarchical care and computer-aided diagnosis. Our two-stage strategy effectively improves the robustness of the model and can be extended to screen for other fetal heart development defects.

Genetically predicted circulating levels of glycine, glutamate, and serotonin in relation to the risks of three major neurodegenerative diseases: A Mendelian randomization analysis.

It has been previously postulated that blood neurotransmitters might affect risks of neurodegenerative diseases. Here, a Mendelian Randomization (MR) study was conducted to explore whether genetically predicted concentrations of glycine, glutamate and serotonin were associated with risks of Alzheimer's disease (AD), Parkinson's disease (PD), and amyotrophic lateral sclerosis (ALS). From three genome-wide association studies of European ancestry, single nucleotide polymorphisms strongly associated with glycine, glutamate and serotonin were selected as genetic instrumental variables. Corresponding summary statistics were also obtained from the latest genome-wide association meta-analyses of AD, PD and ALS. The inverse-variance weighted MR and multiple sensitivity analyses were performed to evaluate causal effects of genetically predicted levels of neurotransmitters on risks of neurodegenerative diseases. The statistical significance threshold was set at P < 0.0056 using the Bonferroni-correction, while 0.0056 < P < 0.05 was considered suggestive evidence for a causal association. There was a causal association of elevated blood glutamate levels with higher AD risks. The odds ratio (OR) of AD was 1.311 [95% confidence interval (CI), 1.087-1.580; P = 0.004] per one standard deviation increase in genetically predicted glutamate concentrations. There was suggestive evidence in support of a protective effect of blood serotonin on AD (OR = 0.607; 95% CI, 0.396-0.932; P = 0.022). Genetically predicted glycine levels were not associated with the risk of AD (OR = 1.145; 95% CI, 0.939-1.396; P = 0.180). Besides, MR analyses indicated no causal roles of three blood neurotransmitters in PD or ALS. In conclusion, the MR study provided evidence supporting the association of elevated blood glutamate levels with higher AD risks and the association of increased blood serotonin levels with lower AD risks. Triangulating evidence across further study designs is still warranted to elucidate the role of blood neurotransmitters in risks of neurodegenerative diseases.

Assessment of causal effects of physical activity on neurodegenerative diseases: A Mendelian randomization study.

BACKGROUND: Physical activity has been hypothesized to play a protective role in neurodegenerative diseases. However, effect estimates previously derived from observational studies were prone to confounding or reverse causation. METHODS: We performed a two-sample Mendelian randomization (MR) analysis to explore the causal association of accelerometer-measured physical activity with 3 common neurodegenerative diseases: Alzheimer's disease (AD), Parkinson's disease (PD), and amyotrophic lateral sclerosis (ALS). We selected genetic instrumental variants reaching genome-wide significance (p < 5 × 10-8) from 2 largest meta-analyses of about 91,100 UK Biobank participants. Summary statistics for AD, PD, and ALS were retrieved from the up-to-date studies in European ancestry led by the international consortia. The random-effect, inverse-variance weighted MR was employed as the primary method, while MR pleiotropy residual sum and outlier (MR-PRESSO), weighted median, and MR-Egger were implemented as sensitivity tests. All statistical analyses were performed using the R programming language (Version 3.6.1; R Foundation for Statistical Computing, Vienna, Austria). RESULTS: Primary MR analysis and replication analysis utilized 5 and 8 instrumental variables, which explained 0.2% and 0.4% variance in physical activity, respectively. In each set, one variant at 17q21 was significantly associated with PD, and MR sensitivity analyses indicated them it as an outlier and source of heterogeneity and pleiotropy. Primary results with the removal of outlier variants suggested odds ratios (ORs) of neurodegenerative diseases per unit increase in objectively measured physical activity were 1.52 for AD (95% confidence interval (95%CI): 0.88-2.63, p = 0.13) and 3.35 for PD (95%CI: 1.32-8.48, p = 0.01), while inconsistent results were shown in the replication set for AD (OR = 1.06, 95%CI: 1.01-1.12, p = 0.02) and PD (OR = 0.99, 95%CI: 0.88-0.12, p = 0.97). Similarly, the beneficial effect of physical activity on ALS (OR = 0.51, 95%CI: 0.29-0.91, p = 0.02) was not confirmed in the replication analysis (OR = 0.96, 95%CI: 0.91-1.02, p = 0.22). CONCLUSION: Genetically predicted physical activity was not robustly associated with risk of neurodegenerative disorders. Triangulating evidence across other studies is necessary in order to elucidate whether enhancing physical activity is an effective approach in preventing the onset of AD, PD, or ALS.

Sleep Duration and Stroke: A Mendelian Randomization Study.

Study Objectives: To clarify the effects of sleep duration on stroke and stroke subtypes, we adopted a Mendelian randomization (MR) approach to evaluate their causal relationship. Methods: A genome-wide association study including 446,118 participants from UK biobank was used to identify instruments for short sleep, long sleep and sleep duration. Summary-level data for all stroke, ischemic stroke, intracerebral hemorrhage, and their subtypes were obtained from meta-analyses conducted by the MEGASTROKE consortium. MR analyses were performed using the inverse-variance-weighted method, weighted median estimator, MR pleiotropy residual sum and outlier (MR-PRESSO) test, and MR-Egger regression. Sensitivity analyses were further performed using leave-one-out analysis, MR-PRESSO global test and Cochran's Q test to verify the robustness of our findings. Results: By two-sample MR, we didn't find causal associations between sleep duration and risk of stroke. However, in the subgroup analysis, we found weak evidence for short sleep in increasing risk of cardio-embolic stroke (odds ratio [OR], 1.33; 95% confidence interval [CI], 1.11-1.60; P = 0.02) and long sleep in increasing risk of large artery stroke [OR, 1.41; 95% CI, 1.02-1.95; P = 0.04]. But the associations were not significant after Bonferroni correction for multiple comparisons. Conclusions: Our study suggests that sleep duration is not causally associated with risk of stroke and its subtypes.

Detailed Echocardiographic Measurements of Individual Chamber in a Chinese Cohort of Hypoplastic Left Heart Syndrome and Comparison with Normal Fetuses via Z-score Modeling.

Hypoplastic left heart syndrome (HLHS) is a rare, but exceptionally serious, congenital heart defect. We aimed to explore the best-fitted Z-score models for individual chamber dimension and to draw a comparison between fetuses with HLHS and the normal Chinese cohort. We made measurements of 1674 healthy fetuses and 79 fetuses with HLHS, undertaking echocardiography. Normal fetal cardiovascular Z-score formulae were established by curve-fitting with 5 algorithmic functions and weighted regression of absolute residuals. Classic linear models were fitted for left ventricular diameter against gestational age, and log-transformed linear-power models-were statistically better for left ventricular length, diameter of left atrium and ascending aorta. Fetuses with HLHS manifested significantly lower Z-score means (≤3.5) for these 4 parameters and the vast majority (∼90%) lay beyond -2. Overall, cardiovascular Z-score equations were reliably constructed in a larger Chinese cohort, and their application should benefit evaluation and diagnosis of HLHS.