Food breakdown during gastric digestion influences gastric emptying, satiety, and nutrient bioaccessibility. However, the engineering design of foods for targeted breakdown has remained a challenge. This is in part due to a lack of data on the breakdown of solid foods during gastric digestion, particularly regarding the influence of simulated peristalsis on the breakdown of solid food particles. In addition, there is a need for standardized model, solid foods for studying food breakdown which could be used to establish a more in-depth understanding of the physical property changes experienced by solid food matrices during in vitro gastric digestion. Such knowledge on the physical property changes of solid foods during in vitro gastric digestion could help to identify the influence of food properties on the mechanisms of particle breakdown during in vitro gastric digestion that includes simulated peristalsis. However, currently available systems for studying food breakdown in the presence of simulated peristalsis include the potential for interactions between food particles as well as for food breakdown due to the direct effect of the simulated peristaltic wave. Thus, there is a need for a new in vitro digestion system that can apply simulated peristaltic waves to single particles of solid food. In this study, progress toward addressing these gaps in the literature was made by assessing the breakdown of almond particles during in vitro digestion both in the presence and in the absence of simulated peristalsis, developing and characterizing standardized, model solid foods with varying breakdown rates during in vitro gastric digestion, and developing a new peristaltic simulator for digestion studies. First, in vitro digestion of almond particles was conducted using a dynamic model (Human Gastric Simulator) and a static model (shaking water bath). This experiment was carried out to gather data on the breakdown of a solid food product during gastric digestion, and to determine whether the observed breakdown could be attributed to the peristaltic motion of the simulated stomach. Results showed that structural breakdown of particles occurred only in the Human Gastric Simulator, as evidenced by a reduction in particle size during the gastric phase of digestion (15.89 ± 0.68 mm2 to 12.19 ± 1.29 mm2, p < 0.05). Fatty acid bioaccessibility at the end of the gastric phase was greater in the Human Gastric Simulator than in the shaking water bath (6.55 ± 0.85% vs. 4.54 ± 0.36%, p < 0.01). Results indicated that the in vitro model of digestion incorporating simulated peristaltic contractions (Human Gastric Simulator) led to breakdown of almond particles during gastric digestion which increased fatty acid bioaccessibility. These results underscore the importance of simulated peristaltic contractions towards achieving realistic breakdown of solid food particles during in vitro digestion. However, due to the inhomogeneous microstructure and particle size distribution of almond particles, it was not possible to determine changes in mechanical properties of almond particles over time during digestion. This pointed to the need for standardized, model foods whose physical properties could be quantified over time during digestion and then related to particle breakdown during digestion experiments that include simulated peristalsis. To meet this research need, standardized, model solid foods were developed based on the macronutrient composition of the Standard American Diet. The model foods were based on hydrogels, which allowed for controllable particle shapes and sizes to be prepared, allowing for assessment of their physical properties using texture analysis, rheology, and fracture property testing. A strong gel version and a weak gel version were created by modulating the pH at which the model foods were produced. Another class of model foods was developed using whey protein hydrogels to serve as a simpler model system. A strong gel and weak gel version of the whey protein model foods were produced by addition of pectin. The whey protein weak gel had the lowest initial hardness of the model foods (6.2 ± 1.0 N) which was significantly (p<0.05) lower than that of the whey protein strong gel (56.2 ± 5.7 N), which had the highest hardness of the four model foods. The breakdown of model foods was assessed during in vitro gastric digestion using texture analysis, dry solids retention ratio analysis, and modeling of moisture diffusion into the model food matrix. The standard diet weak gel had the fastest softening half-time of the model foods (58 ± 5 min), which was significantly (p<0.05) lower than that of the whey protein strong gel (775 ± 82 min) which had the longest softening half-time of the model foods. Fracture properties of the model foods were assessed using wire cutting and knife cutting. Toughness, yield stress, and stiffness were significantly influenced by digestion time (p < 0.01). Rheological properties of model foods were assessed using parallel plate oscillatory rheology, and it was found that the storage modulus (G’), loss modulus (G’’), and tan δ were significantly influenced by digestion time (p < 0.01). One unexpected finding from the study of fracture properties was that the apparent toughness of the whey protein strong gel increased during the first three hours of in vitro gastric digestion but decreased after 24 hours of in vitro gastric digestion. The initial increase in toughness of this model food was attributed to a loss of brittleness due to moisture uptake, with the eventual decrease in apparent toughness due to the weakening of the food matrix by acid and enzymatic hydrolysis. This suggests that competing mechanisms operating on different time scales can lead to non-monotonic changes in food fracture properties during digestion, specifically, rapid moisture uptake can lead to increased toughness, whereas slower matrix hydrolysis eventually leads to decreased toughness. Taken together, these findings show that physical property changes in soft solid foods during in vitro gastric digestion are complex. After characterizing the changes in textural, fracture, and rheological properties of the model foods over time during in vitro gastric digestion, it was desired to relate these properties to the breakdown mechanisms experienced by the model foods during in vitro gastric digestion that included simulated peristalsis. However, existing dynamic digestion systems apply simulated peristaltic contractions to numerous food particles simultaneously, meaning that particle/particle grinding could contribute to food breakdown in addition to the direct effect of the peristaltic wave. To address the need for a system that can apply simulated peristaltic contractions to food particles one at a time, a novel peristaltic simulator was designed and constructed. The new peristaltic simulator can apply simulated peristaltic contractions one at a time to a single food particle, and also allows for up to 12 digestion modules to be used simultaneously. This >10-fold improvement in throughput relative to previous state-of-the-art systems could allow for fast screening of food materials for health-promoting properties. The novel simulator was characterized using quantitative video analysis and computational fluid dynamics (CFD) simulations, the results of which were compared to physiological data on the motility of the human stomach and intestine. This new digestion simulator could help future researchers identify the mechanisms leading to food breakdown by studying the effect of digesta rheology or peristaltic wave dynamics on food breakdown. One important characteristic of the novel peristaltic simulator is that it permits simulated peristaltic contractions of adjustable occlusion to be applied one at a time to individual food particles allowing their breakdown to be studied in more detail. The model foods developed with varying breakdown rates during static in vitro digestion were then subjected to static in vitro digestion for varying lengths of time followed by digestion using either the Human Gastric Simulator (HGS) or the novel peristaltic simulator, systems which include simulated peristalsis. The HGS was used to study the breakdown of particles of model, solid foods under physiologically representative conditions which included the potential for breakdown due to the direct effect of the simulated peristaltic wave as well as due to particle/particle grinding. The novel peristaltic simulator was used to isolate the effect of the simulated peristaltic wave without the potential for breakdown due to particle/particle grinding. Results showed that the number and sizes of particles were significantly (p < 0.05) influenced by the type of model food and in vitro digestion time. Results from this study were used to establish the hypothesized breakdown mechanisms for the standardized, model, solid foods during in vitro digestion that includes simulated peristalsis. Specifically, the standard diet strong gel broke down by erosion as well as chipping, whereas the standard diet weak gel broke down by erosion, chipping, and fragmentation. The whey protein strong gel broke down by erosion with only minor breakdown overall, while the whey protein weak gel broke down by erosion, chipping, and fragmentation. Experiments using the peristaltic simulator suggested that the mechanisms of particle breakdown of the model, solid foods were influenced by the number of simulated peristaltic contractions that were applied to the particle. For example, the standard diet strong gel broke down by erosion and chipping until ca. 30 simulated peristaltic contractions had been applied, after which the breakdown could be ascribed to erosion, chipping, and fragmentation. These results demonstrated that the mechanisms which lead to particle breakdown during in vitro digestion that includes simulated peristalsis may be dynamic in time, so that foods which initially break down only by erosion or erosion and chipping may be weakened by the progressive application of contractions, leading to breakdown at later times due to erosion, chipping, and fragmentation. Overall, the model foods, which were shown to fall into distinct classes in the Food Breakdown Classification System (FBCS), experienced differing hypothesized breakdown mechanisms which were related to their initial properties and kinetics of texture change. It was determined that that food particles with hardness < 6 N may break down by erosion, chipping, and fragmentation at the onset of simulated peristalsis but particles with hardness > 9 N may break down initially by erosion but transition to breakdown by erosion and chipping at later digestion times as the particle softens due to contact with simulated gastric fluid. Collectively, the data gathered in these studies show that particle breakdown of solid foods during in vitro gastric digestion was driven by simulated peristaltic contractions, that the physical properties of solid foods influenced their breakdown rates during in vitro digestion, and that the progressive application of simulated peristaltic contractions can lead to transitions in the breakdown mechanisms of solid food particles. These findings could help future researchers develop foods with targeted rates of particle breakdown during gastric digestion, a potential means to modulate gastric emptying rates in vivo. The system of standardized model, solid foods and novel peristaltic simulator could serve as tools to help future researchers develop foods with targeted breakdown rates for controlled bioaccessibility or satiety, for example, by linking the initial values of food physical properties and their changes during in vitro gastric digestion to their expected mechanisms of breakdown during gastric digestion.

Chickpea (Cicer arietinum L.) is an agriculturally-important legume crop that is an excellent source of proteins, fiber, and minerals. Therefore, developing chickpea-based snacks could add value to chickpeas, as well as provide consumers with snack products rich in protein and fiber. As two main components, there are around 24% protein and 54% carbohydrate in chickpea that could be influenced by processing methods. Many studies have focused on the effect of processing on in vitro digestibility of starch and nutrient availability in legumes. Previous studies have found that the water uptake during digestion plays an important role in the breakdown of starch-based food matrices. However, the interaction between particle size, water content, and modifications in nutrient release and digestibility is not clear. The objective of this project was to investigate the effect of particle size and water content on particle breakdown and starch digestibility of chickpea-based snacks during dynamic in vitro digestion.To achieve this goal, chickpeas puree and cracker were produced with different particle size: fine puree, fine cracker, coarse puree, coarse cracker. For all treatments, chickpea samples were mixed with simulated saliva for 30 s and were incubated in the Human Gastric Simulator (HGS) for up to 180 min. Gastric digesta were removed from the bottom of HGS every 30 min to be analyzed and sample aliquots were also incubated in a shaking water bath (37℃ at 100 rpm) with simulated intestinal juice for up to 180 min. Intestinal digesta were removed from simulated intestinal juice every 30 min up to 180 min to be analyzed. For whole gastric digesta, pH, moisture content, and gastric emptying were measured. Gastric digesta were then separated into 3 phases: liquid, suspended solid, and solid phase. For the suspended solid phase, particle size was measured using laser diffraction, while image analysis was used to measure the particle size and number of particles in the solid phase. The starch hydrolysis and protein hydrolysis were measured in all three phases through analysis of reducing sugars and free amino groups, respectively. It was found that both the initial particle size and drying influenced the breakdown rate and the digestibility of starch and protein in chickpea snacks. Across the four treatments, the pH decreased from 3.38-4.00 after 30 min and to 1.61-1.14 after 180 min gastric digestion. The pH of the fine puree was the highest from 60-150 min compared with the other treatments, as it had the highest buffering capacity of the four treatments. For the gastric emptying of dry matter, the emptying rate (k) was significantly (p<0.05) higher for fine puree compared to the other treatments, which may be due to the higher pH and lower water holding capacity observed in the fine puree. The particle size of the solid phase was quantified through the x50, or median particle area. The x50 of fine puree was constant during gastric digestion due to the smallest particle size, while fine cracker and coarse cracker rapidly decreased in size from 0-30 min due to changes in the product matrix that occur during drying. The coarse puree had the largest particle size at all digestion times, due to the remaining large pieces of skin that were not broken down during milling. The hydrolysis of starch and protein in the liquid phase during gastric digestion had a similar trend: fine cracker>coarse cracker>fine puree>coarse puree, which was impacted by both initial particle size and drying. The reducing sugars and free amino groups released from fine puree were higher than coarse puree, and fine cracker was higher than coarse cracker due to the influence of initial particle size. Similarly, the water content also impacted starch and protein hydrolysis in the chickpea snacks, crackers showed higher protein and starch hydrolysis than puree. Due to their high protein content and potentially low glycemic index, developing chickpea-based snacks could result in nutritious, slowly-digestible high protein snacks. Therefore, this study is essential to understand how the properties and processing of chickpeas, such as particle size and moisture content, impact their breakdown and macronutrient digestibility. This information will help to develop chickpea-based snacks with specific functional properties to control starch and protein digestion.

Nondestructive imaging combined with recent advances in data processing techniques allows for minimally invasive, high-throughput, time-series analysis in both food and agriculture. The flexibility of neural network architectures allows for prediction of continuous output metrics, which may be valuable in analysis of both food and agriculture. In this work, nondestructive image analysis was used to demonstrate the relevance of image data to physical properties in food and agricultural systems as well as the potential for predictive models to increase the efficiency of quantitative data analysis.Apples undergoing in vitro gastric digestion were used as a model food system, and X-ray micro-computed tomography data were shown to relate to peak force of apple tissue during compression. This relationship was probed further using a deep learning approach, where the compression curves of apple tissue during in vitro gastric digestion were predicted using a regression convolutional neural network (CNN) model. Results under cross-validation demonstrated strong accordance between predicted and measured compression curves, with an R2 of 0.939 and RMSE of 4.36 N across measured force values. This relationship declined in samples from a holdout set, with an RMSE of 14.3 N, although this result was influenced strongly by the incubation medium tested (water vs. gastric juice). Within the agricultural domain, the task of nondestructive yield estimation was demonstrated in vineyards. Images of grapevines collected using proximal sensing were associated with yield measured at harvest using a commercial yield monitor, allowing for a dataset of 23,581 yield measurements predicted using 164,699 images to be collected. Three deep learning architectures were used to predict the measured yield values from the images: object detection, CNN regression, and a transformer regression network. Regression-based architectures were used to eliminate the bottleneck of hand-labeling images. Results demonstrated that regression methods performed comparably to object detection methods without the need for hand labeling. Grape yield from within a training set, was correlated with model output at a mean absolute percent error of 6.4% when aggregated into 10 m regions using a transformer model. This study also demonstrated performance on a representative holdout set, with an error of 18% obtained using the same model and conditions. Overall, this study demonstrated the potential of deep learning combined with nondestructive imaging for quantitative analysis of food and agricultural systems. Results demonstrated that image data can be related to mechanical properties of food materials as well as yield in an agricultural environment.