Amaranth (Amarantus caudatus), quinoa (Chenopodium quinoa) and kañiwa (Chenopodium pallidicaule) are Andean grains that are gaining interest as nutritious gluten-free alternatives to conventional cereals. Near infrared (NIR) imaging was applied to extrudates containing 20, 35% and 50% amaranth, quinoa and kañiwa in order to study the spatial distribution of fibre and protein along the cross-sectional area. The results were contrasted with existing physical measurements (e.g., sectional expansion, stiffness) and textural data obtained from sensory profiling and temporal studies (i.e., temporal dominance of sensation, TDS). Score distribution in PCA plots was directly associated to fibre (PC1) and protein (PC2) due to spectral wavelength specificity (fibre: 1028 nm; protein: 1470 nm). Partial Least Squares regression model (PLS) showed that evenly distributed protein structures are strongly linked to desirable TDS textural properties such as crispiness and crunchiness, while protein clumps were linked to undesirable properties such as roughness. In contrast, fibre was found to reduce roughness. PLS could not explain accurately changes in physical attributes, and sensory data from profiling tests had to be omitted from computing due to lack of fit. This study shows that NIR hyperspectra imaging could help elucidate the chemical background of physical and particularly temporal dominant attributes.