RGB pattern of images allows rapid and efficient prediction of antioxidant potential in Calycophyllum spruceanum barks

Predictive modeling of antioxidant activity in Syzygium malaccense leaf extracts using image processing and machine learning

S. malaccense, from the Myrtaceae family, is used in traditional medicine and is rich in flavonoids and phenolic compounds. This study evaluated the antioxidant potential of S. malaccense leaf extracts and their fractions using DPPH and ABTS radical scavenging assays, Ferric Reducing Antioxidant Power (FRAP), and total phenolic content. Spectroscopic methods were used, and greyscale tones from the RGB channels of assay images were analyzed through machine learning (ML) models such as SVM, decision tree, Random Forest (RF), XGBOOST, LightGBM, and CatBoost. The performance of these models was assessed using determination coefficients (R2) and root mean square error (RMSE). XGBOOST and RF were the best performers, with R2 values ranging from 88.65 to 99.35% for training data and 60.12-95.50% for test data. GLM analysis showed that acetate solvent resulted in the highest FRAP values, while hexane had the lowest. Ethanol extraction yielded the highest ABTS values, and dichloromethane was best for DPPH. These modeling approaches using GLM, images, and ML algorithms show promise for measuring the antioxidant properties of plants.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13197-024-06073-2.

An eco-friendly approach for analysing sugars, minerals, and colour in brown sugar using digital image processing and machine learning

Brown sugar is a natural sweetener obtained by thermal processing, with interesting nutritional characteristics. However, it has significant sensory variability, which directly affects product quality and consumer choice. Therefore, developing rapid methods for its quality control is desirable. This work proposes a fast, environmentally friendly, and accurate method for the simultaneous analysis of sucrose, reducing sugars, minerals and ICUMSA colour in brown sugar, using an innovative strategy that combines digital image processing acquired by smartphone cell with machine learning. Data extracted from the digital images, as well as experimentally determined contents of the physicochemical characteristics and elemental profile were the variables adopted for building predictive regression models by applying the kNN algorithm. The models achieved the highest predictive capacity for the Ca, ICUMSA colour, Fe and Zn, with coefficients of determination (R2) ≥ 92.33 %. Lower R2 values were observed for sucrose (81.16 %), reducing sugars (85.67 %), Mn (83.36 %) and Mg (86.97 %). Low data dispersion was found for all the predictive models generated (RMSE < 0.235). The AGREE Metric assessed the green profile and determined that the proposed approach is superior in relation to conventional methods because it avoids the use of solvents and toxic reagents, consumes minimal energy, produces no toxic waste, and is safer for analysts. The combination of digital image processing (DIP) and the kNN algorithm provides a fast, non-invasive and sustainable analytical approach. It streamlines and improves quality control of brown sugar, enabling the production of sweeteners that meet consumer demands and industry standards.

Chemical profile and biological properties of the Piper corcovadense C.DC. essential oil

The essential oil from Piper corcovadense D.DC. (EOPc), an important plant belonging to the Piperaceae family, which is commonly found in the northern region of Brazil and poorly explored scientifically, was used in this study. Thus, the EOPc was characterized chemically by Gas Chromatography/Mass Spectrometry (GC/MS) and the antioxidant and antimicrobial activities and their potential effects on cutaneous melanoma (SK-MEL-28) and healthy peripheral blood mononuclear (PBMC) cells were determined. The major compounds identified in the EOPc were: trans-sesquisabinene hydrate, trans-caryophyllene, β-pinene, trans-β-farnesene, 14-hydroxycaryophyllene, limonene and p-cymene. The EOPc demonstrated antioxidant activity as evaluated by Folin-Ciocalteu reagent (FC) reducing capacity, DPPH, and ABTS methods. The values found were respectively 5.41 ± 0.17 mg GAE mL-1 (GAE: Gallic acid equivalent), 2.88 ± 0.17 µmol TE mL-1 (TE: Trolox equivalent) and 6.26 ± 0.02 µmol TE mL-1. The minimum inhibitory concentration (MIC) and minimum bactericidal concentration (MBC) were determined for different bacterial strains. The EOPc at a concentration of 2.61 µg mL-1 exhibited both bactericidal and bacteriostatic properties against Escherichia coli. The EOPc showed potential antitumor activity as it reduced the cell viability of human cutaneous melanoma cells SK-MEL-28. Besides, the EOPc did not exhibit cytotoxic activity against healthy PBMCs, indicating that it does not harm healthy cells at the tested concentrations. The EOPc increased the levels of ROS at concentrations of 250 µg mL-1. The EOPc also did not stimulate the mobilization of endogenous antioxidant defenses, as assessed by total thiol (PSH) and non-protein thiols (NPSH). Thus, the study suggests that the EOPc has antioxidant and antimicrobial properties due to the presence of specific compounds. It also exhibits antitumor potential against cutaneous melanoma cells while showing no cytotoxicity to healthy PBMCs. It directly influenced ROS levels at the highest tested concentration in the cells, suggesting an antitumor effect related to the intrinsic apoptosis pathway. Nevertheless, while the study has initial findings, the results are promising and indicate an attractive biological potential of P. corcovadense, mainly in human cutaneous melanoma cells.

Phenotyping Key Fruit Quality Traits in Olive Using RGB Images and Back Propagation Neural Networks

To predict oil and phenol concentrations in olive fruit, the combination of back propagation neural networks (BPNNs) and contact-less plant phenotyping techniques was employed to retrieve RGB image-based digital proxies of oil and phenol concentrations. Fruits of cultivars (×3) differing in ripening time were sampled (~10-day interval, ×2 years), pictured and analyzed for phenol and oil concentrations. Prior to this, fruit samples were pictured and images were segmented to extract the red (R), green (G), and blue (B) mean pixel values that were rearranged in 35 RGB-based colorimetric indexes. Three BPNNs were designed using as input variables (a) the original 35 RGB indexes, (b) the scores of principal components after a principal component analysis (PCA) pre-processing of those indexes, and (c) a reduced number (28) of the RGB indexes achieved after a sparse PCA. The results show that the predictions reached the highest mean R² values ranging from 0.87 to 0.95 (oil) and from 0.81 to 0.90 (phenols) across the BPNNs. In addition to the R², other performance metrics were calculated (root mean squared error and mean absolute error) and combined into a general performance indicator (GPI). The resulting rank of the GPI suggests that a BPNN with a specific topology might be designed for cultivars grouped according to their ripening period. The present study documented that an RGB-based image phenotyping can effectively predict key quality traits in olive fruit supporting the developing olive sector within a digital agriculture domain.

Health Evaluation and Fault Diagnosis of Medical Imaging Equipment Based on Neural Network Algorithm

In recent years, high-precision medical equipment, especially large-scale medical imaging equipment, is usually composed of circuit, water, light, and other structures. Its structure is cumbersome and complex, so it is difficult to detect and diagnose the health status of medical imaging equipment. Based on the vibration signal of mechanical equipment, a PLSR-DNN hybrid network model for health prediction of medical equipment is proposed by using partial least squares regression (PLSR) algorithm and deep neural networks (DNNs). At the same time, in the diagnosis of medical imaging equipment fault, the paper proposes to use rough set to screen the fault factors and then use BP neural network to classify and identify the fault and analyzes the practical application effect of the two technologies. The results show that the PLSR-DNN hybrid network model for health prediction of medical imaging equipment is basically consistent with the actual health value of medical equipment; medical imaging equipment fault diagnosis technology is based on rough set and BP neural network. In the test set, the sensitivity, specificity, and accuracy of medical imaging equipment fault identification are 75.0%, 83.3%, and 85.0%. The above results show that the proposed health prediction method and fault diagnosis method of medical imaging equipment have good performance in health prediction and fault diagnosis of medical equipment.