Visible and near-infrared spectral imaging combined with robust regression for predicting firmness, fatness, and compositional properties of fresh pork bellies

Belly is a widely consumed pork product with very variable properties. Meat industry needs real-time quality assessment for maintaining superior pork quality throughout the production. This study explores the potential of using visible and near-infrared (VNIR,386-1015 nm) spectral imaging for predicting firmness, fatness and chemical compositional properties in pork belly samples, offering robust spectral calibrations. A total of 182 samples with wide variations in firmness and compositional properties were analysed using common laboratory analyses, whereas spectral images were acquired with a VNIR spectral imaging system. Exploratory analysis of the studied properties was performed, followed by a robust regression approach called iterative reweighted partial least-squares regression to model and predict these belly properties. The models were also used to generate spatial maps of predicted chemical compositional properties. Chemical properties such as fat, dry matter, protein, ashes, iodine value, along with firmness measures as flop distance and angle, were predicted with excellent, very good and fair models, with a ratio prediction of standard deviation (RPD) of 4.93, 3.91, 2.58, 2.54, 2.41, 2.53 and 2.51 respectively. The methodology developed in this study showed that a short wavelength spectral imaging system can yield promising results, being a potential benefit for the pork industry in automating the analysis of fresh pork belly samples. VNIR spectral imaging emerges as a non-destructive method for pork belly characterization, guiding process optimization and marketing strategies. Moreover, future research can explore advanced data analytics approaches such as deep learning to facilitate the integration of spectral and spatial information in joint modelling.

Automatic Quality Assessment of Pork Belly via Deep Learning and Ultrasound Imaging

Pork belly, prized for its unique flavor and texture, is often overlooked in breeding programs that prioritize lean meat production. The quality of pork belly is determined by the number and distribution of muscle and fat layers. This study aimed to assess the number of pork belly layers using deep learning techniques. Initially, semantic segmentation was considered, but the intersection over union (IoU) scores for the segmented parts were below 70%, which is insufficient for practical application. Consequently, the focus shifted to image classification methods. Based on the number of fat and muscle layers, a dataset was categorized into three groups: three layers (n = 1811), five layers (n = 1294), and seven layers (n = 879). Drawing upon established model architectures, the initial model was refined for the task of learning and predicting layer traits from B-ultrasound images of pork belly. After a thorough evaluation of various performance metrics, the ResNet18 model emerged as the most effective, achieving a remarkable training set accuracy of 99.99% and a validation set accuracy of 96.22%, with corresponding loss values of 0.1478 and 0.1976. The robustness of the model was confirmed through three interpretable analysis methods, including grad-CAM, ensuring its reliability. Furthermore, the model was successfully deployed in a local setting to process B-ultrasound video frames in real time, consistently identifying the pork belly layer count with a confidence level exceeding 70%. By employing a scoring system with 100 points as the threshold, the number of pork belly layers in vivo was categorized into superior and inferior grades. This innovative system offers immediate decision-making support for breeding determinations and presents a highly efficient and precise method for assessment of pork belly layers.

Can In-Line Iodine Value Predictions (NitFomTM) Be Used for Early Classification of Pork Belly Firmness?

Commercial technologies for assessing meat quality may be useful for performing early in-line belly firmness classification. This study used 207 pork carcasses to measure predicted iodine value (IV) at the clear plate region of the carcass with an in-line near-infrared probe (NitFom^TM), calculated IV of belly fat using wet chemistry methods, determined the belly bend angle (an objective method to measure belly firmness), and took dimensional belly measurements. A regression analysis revealed that NitFom^TM predicted IV (R² = 0.40) and belly fat calculated IV (R² = 0.52) separately contributed to the partial variation of belly bend angle. By testing different NitFom^TM IV classification thresholds, classifying soft bellies in the 15th percentile resulted in 5.31% false negatives, 5.31% false positives, and 89.38% correctly classified soft and firm bellies. Similar results were observed when the classification was based on belly fat IV calculated from chemically analyzed fatty acid composition. By reducing the level of stringency on the percentile of the classification threshold, an increase in false positives and decrease in false negatives was observed. This study suggests the IV predicted using the NitFom^TM may be useful for early in-line presorting of carcasses based on expected belly firmness, which could optimize profitability by allocating carcasses to specific cutout specifications.

Prediction of carcass composition and meat and fat quality using sensing technologies: A review

Consumer demand for high-quality healthy food is increasing,  thus meat processors require the means toassess these rapidly, accurately, and inexpensively. Traditional methods forquality assessments are time-consuming, expensive, invasive, and have potentialto negatively impact the environment. Consequently, emphasis has been put onfinding non-destructive, fast, and accurate technologies for productcomposition and quality evaluation. Research in this area is advancing rapidlythrough recent developments in the areas of portability, accuracy, and machinelearning. The present review, therefore, critically evaluates and summarizes developmentsof popular non-invasive technologies (i.e., from imaging to spectroscopicsensing technologies) for estimating beef, pork, and lamb composition andquality, which will hopefully assist in the implementation of thesetechnologies for rapid evaluation/real-timegrading of livestock products in the nearfuture.