Search

This study was conducted to investigate the effects of different mixture ratios of sorghum (S) and cluster bean (CB) on fermentation parameters, nutritional composition, and gas-methane production of mixture silages. Five different mixture ratios (100%S, 75%S+25%CB, 50%S+50CB, 25%S+75%CB and 100%CB) were used in this study. Effects of mixture ratios on entire parameters were highly significant (p ≤ 0.01). Decreasing ADF and NDF ratios and increasing crude protein, ash and oil contents and condensed tannin levels were observed with increasing cluster bean ratios of the mixtures. However, cluster beans reduced WSC and lactic acid and increased pH, butyric acid and ethanol contents. The lowest Fleig score was obtained from cluster bean silage (100%CB). The lowest gas-methane production was obtained from cluster bean silage and the greatest ME and NEL values were obtained from sorghum silage (100%S). The best OMD values were obtained from sorghum silage and 75%S +25%CB mixture silage. The present findings revealed that a 50%S+50%CB mixture yielded high nutritional composition and good fermentation parameters. [ABSTRACT FROM AUTHOR]

Background: Predicting disease classification in youth with a first episode of demyelination is feasible in some but not all cases. Machine learning algorithms can provide highly accurate predictions of outcomes directly from images, removing a major bottleneck in the speed with which predictions can be made. Objectives: To use visual metrics to predict disease class in youth with demyelination. Methods: Standardized, prospectively collected clinical and visual data at disease onset from 224 pediatric subjects, classified using consensus definitions of demyelinating disorders and serum antibody testing for myelin oligodendrocyte glycoprotein (MOG) and aquaporin 4 (healthy control = 72, multiple sclerosis = 69, anti-MOG = 18, neuromyelitis optica spectrum disorder [NMOSD] = 10, monophasic acquired demyelinating syndromes [MonoADSs] = 55) were recruited through the Demyelinating Disorders Program at The Hospital for Sick Children (Toronto, Ontario) and University of Calgary (2010-2019). We used 9 supervised machine learning algorithms (random forest classifier, CLS bagging classifier, AdaBoost classifier, XGBoost classifier, CLS logistic regression, MLP classifier, CLS SVM, decision tree, and Kneighbors classifier). Input variables were optical coherence tomography (OCT) parameters including retinal nerve fiber layer (RNFL) and ganglion cell inner plexiform layer thickness (GCIPL), low-contrast visual acuity (LCVA), and color vision for each eye. Separate sets of analyses were run with balanced (n set arbitrarily at 44, oversampling and undersampling used appropriately) and unbalanced data. Results: In the analysis of the balanced data set, the random forest classifier (accuracy [a] = 80%, recall [r] = 50%, and precision [p] = 99%), XGBoost classifier (a = 80%, r = 50%, p = 80%), and decision tree (a = 90%, r = 99%, p = 90%) algorithms yielded the highest accuracy, recall, and precision levels for each disease class using the combination of RNFL, GCIPL, LCVA, and color vision. Analysis of the unbalanced data set showed lower overall levels of predictive accuracy (60%-90%) for each class using the same algorithms. Use of the OCT data alone yielded lower predictive accuracy in both balanced and unbalanced analyses. Conclusions: Machine learning algorithms used on combined structural and functional visual metrics may classify youth with demyelinating disorders. Implementation of artificial neural networks using raw OCT data and images as input are underway. [ABSTRACT FROM AUTHOR]