Physicochemical properties of extrudate-based flakes from whole banana flour and rice flour blends

Abstract This study aimed to understand how adding whole-banana flour (WBF) affected the physicochemical properties of extrudates intended for the development of breakfast flakes. The parameters: WBF addition (50% to 70%, dry basis), feed moisture (27% to 33%, wet basis), and barrel temperature in the last zone of the extruder (80 ºC to 100 °C) were varied according to a 23 factorial design with augmented center points. The addition of WBF linearly affected all the physical and chemical properties evaluated and showed a synergistic effect with barrel temperature for changes in hardness and carbohydrate content. WBF additions above 64.9%, adjusted to feed moisture between 29.6% to 32.6%, and thermomechanically cooked at barrel temperatures above 87 °C, consumed 216 to 257 kJ/kg of mechanical energy, which produced flakes with a hardness between 160 and 180 N. This work could contribute to the incorporation of WBF for developing more nutritious breakfast cereals, with a fiber content greater than 7.2 g/100 g. HIGHLIGHTS Whole banana flour is a natural source of dietary fiber, vitamins, and minerals, which works well in the development of breakfast flakes Flakes made from whole banana and rice flour are naturally gluten-free, making them an excellent choice for those with celiac disease or gluten intolerance Green-extraction technique for the analysis of antioxidant activity Whole banana flour positively affects the physicochemical properties of extruded flakes

to wheat flour continues as well as to provide sustainable by-product management, other forms of green banana flour consumption should be attempted.The peel contains healthy compounds, such as minerals and phenolic compounds (Pico et al., 2019), and increases lipids, proteins, and dietary fiber (Khoozani et al., 2019).This characteristic composition, together with the resistant starch provided by the pulp (Garcia-Valle et al., 2020), makes WBF a functional ingredient that stimulates intestinal peristalsis, fermentative capacity in the digestive tract, bowel evacuation, and a low glycemic index (Patiño-Rodríguez et al., 2019).
Breakfast flakes are conventionally produced by hydrothermal treatment or extrusion following the flaking of pellets, usually using cereal flours such as wheat, corn, and rice (Fast & Caldwell, 2000).Several breakfast cereal-like products have been formulated by partially replacing these cereal flours with other sources such as black gram (flaxseed) (Rani et al., 2020), cowpea bean (Marengo et al., 2017), and banana flour (Borah et al., 2016;Kaur et al., 2015).However, the inclusion of fiber-rich flour from non-conventional sources affects pellet texture and extrudate solubility; therefore, appropriate incorporation must be determined so that it does not significantly affect these properties.There is limited information in the literature on the use of WBF to formulate food products.Kaur et al. (2015) replaced 0 to 30% banana pulp flour in the development of a breakfast based on corn, and Borah et al. (2016) used 10% to 30% seeded banana flour during the formulation of breakfast cereal based on low-amylose rice flour and carambola pomace.Higher inclusions of could ensure more nutritive final products, with a prebiotic function, due to the fiber content effect in the gastrointestinal tract (Patiño-Rodríguez et al., 2019).Khoozani et al. (2020) developed bread by replacing wheat flour with WBF in the range of 10% to 30%, reaching 5.92% resistant starch in bread formulated with 30% of WBF.
Other important independent variables during extrudate production are feed moisture and barrel temperature profiles (Leonard et al., 2020;Masli et al., 2018).This moisture range was sufficient for pellet formation.For whole wheat flour, pellet moisture levels must be between 22% and 24%.The necessary heat input to the barrel must ensure dough cooking, considering that the zone temperature near the feeding point must be sufficiently low to prevent the start of starch conversion (Fast & Caldwell, 2000).Starch conversion monitoring during the extrusion process is carried out through specific mechanical energy (SME) input, whereas for extrudates (extruded flour suspensions), it is performed by measuring the water solubility index (WSI) and paste viscosity parameters.Accordingly, it was proposed to explore feeding blend moisture in the range of 27% to 33% and barrel temperatures in the last zone of 80 °C to 100 °C.These changes can affect the texture, solubility, and chemical composition of extrudates.The texture of breakfast flakes must be harder than that of extruded snacks because, as they are gradually hydrated during consumption with milk, they must retain the desired crunchy texture for a longer time.The final texture of flakes depends on the degree of pellet flattening, which is fitted by the roller rotation speed and roll gap, and its subsequent drying (Fast & Caldwell, 2000).The hardness (HD) of the bulk flake portion is the usual parameter for determining the instrumental texture of breakfast cereals (Leonard et al., 2020).Thus, this study aimed to evaluate the effects of WBF addition, feed moisture, and barrel temperature in the last zone of the extruder on starch conversion during the extrusion process (SME), starch conversion on flakes (WSI and paste viscosity parameters), HD, and proximal chemical composition of extruded ready-to-eat whole-banana breakfast flakes.In addition, the optimal region was determined using multiple desirable characteristics of the final product.

Materials
White rice grains were acquired from Fumacense Food Ltd. (Santa Catarina, Brazil).Banana fruits ("Terra" variety, genotype AAB) in the third stage of ripening were acquired from a central supply market (CEASA, Rio de Janeiro, Brazil) according to the maturation scale described by Gomes et al. (2013).The banana fruits were washed, immersed in chlorinated water (50 ppm/10 min), rinsed, and drained (3 min).The fruits were then sliced approximately 5 mm thick, submerged in boiling water (3 min), slipped, placed on perforated trays, and dried for 20 h in a Hauber oven (DMS-G-EG, Santa Catarina, Brazil) at 60 °C.

Preparation of flours
White rice grains and dehydrated whole-banana slices were ground under the same conditions in a knifehammer mill (TREU, 1188 model, São Paulo, Brazil), in which a 1.0 mm aperture screening plate was coupled.The milled products, WBF and RF were packed in low-density polyethylene (LDPE) bags and stored until further use.

Particle-size distribution
Particle-size distribution of the raw flours was determined by sieving analysis (American Society of Agricultural and Biological Engineers, 2008).The sample (100 g) was segregated through a set of six standard stainless-steel sieves (297, 212, 149, 106, 90, and 75 μm) and a pan (Newark, USA), using a ROTAP RX-29-10 (WS Tyler, St Albans, USA) for 10 min.

Preparation of feeding blends
The raw flours were blended in proportions according to the experimental design (Table 1).For each trial, blends of 500 g were manually prepared in LDPE bags.A water-appropriate amount (W, g) was added to adjust the moisture content according to Equation 1.
where B is the blend's mass to be moistened (500 g), M 0 is the blend's initial moisture, determined in duplicate by the oven method at 105 °C for 4 h (Association of Official Analytical Chemists, 2005), and M f is the blend's final moisture (g/100 g, wet basis).

Extrusion process
The moistened blends were processed using a single-screw extruder 19/20 DN (Brabender, Duisburg, Germany) equipped with an internal grooved barrel, three electric heating zones, and an L/D ratio of 20.A standard screw attached to the barrel had a compression ratio of 3:1.A 3.0 mm diameter single cylindrical hole die was attached at the output of the third heating zone.The blends were fed using a single-screw volumetric feeder (Brabender, Duisburg, Germany).The feeding rate was set at ~ 4 kg/h.Extruder operational variables such as screw speed (120 rpm), barrel temperature profile (zone 1: 50 °C; zone 2: 70 °C; zone 3: according to the experimental design, Table 1), and torque values (N m) were monitored using WINEXT software, v 4.4.0 (Brabender, Duisburg, Germany).Once the process reached steady-state conditions for each trial, extrudates were collected.Mass flow at the beginning and end of each collection was measured in duplicate by collecting the amount of extrudate for 30 s.

Starch conversion during the extrusion process
Specific mechanical energy (SME) was calculated according to Equation 2.
where T is the torque (N m) recorded at the time of collection, SS is screw speed which was kept constant at 120 rpm, n is the number of screws (n = 1), and Q is extrudate mass flow..

Preparation of flakes and extruded flours
In the intermediate stage of collection, the extrudate stream was coupled to a CL 300 roll laminator (G.Paniz, Rio Grande do Sul, Brazil), and laminated extrudates were cut into strips approximately 40 cm in length.The strips were then manually cut to produce square-shaped flakes of approximately 10.0 mm per side.Moist flakes were dried at 60 °C under the same conditions to produce dehydrated whole banana slices.The dried flakes were stored in polyethylene bags at room temperature for further analysis.The dried flake portion of each trial was ground in a disc mill LM3600 (Perten Instruments AB, Huddinge, Sweden) set to aperture 2.Then, the ground products were powdered in a hammer-mill LM3100 (Perten Instruments AB, Huddinge, Sweden) fitted with a 0.8 mm sieve aperture to reach a desirable particle size for analysis.The extruded flours were packed in LDPE bags and stored at room temperature for further analysis.

Hardness (HD)
The HD of the extruded flakes was determined by a compression test using a texture analyzer (TA.TXplus, Stable Micro Systems, UK), in which a 30 kg load cell and a 50 mm diameter compression plunger were coupled.Dry flakes were poured into a cylindrical container with a diameter of 52 mm and a height of approximately 30 mm (15 g sample).The compression accessory descended at a speed of 2 mm/s and deformed the flake bed to 30% of its initial height.After the test, the compression plunger ascended at 10 mm/s.For each trial, 15 repetitions were performed, and hardness was expressed as the maximum force (N) along the force-distance curve.

Starch conversion in the raw and extruded flours
Raw and extruded flours were sieved and particles between 106 and 212 μm were used for starch conversion analyses.

Water solubility index (WSI)
Suspension preparation and solubilization processes were performed according to the methodology described by Vargas-Solórzano et al. (2014).

Paste viscosity
The paste viscosity profiles were measured using a Rapid Viscosity Analyzer (RVA) series 4 (Newport Scientific Pty Ltd., Warriewood, Australia).The suspensions were prepared by adding 3 g of the sample (14% moisture) to 25 g of distilled water in an RVA aluminum cup.The paddle rotating speed was 160 rpm and the time-temperature profile started at 25 °C for 2 min, then heated to 95 °C, maintained at that temperature for 3 min, and then cooled to 25 °C, completing the test within 20 min.The heating and cooling stages were performed at 14 °C/min.The determined paste viscosity parameters were peak viscosity at 95 °C (PV) and setback viscosity (SB).

Proximal chemical composition
The proximal composition and dietary fiber content of raw and extruded flours were analyzed according to the official methods of the Association of Official Analytical Chemists (2005).The moisture and ash contents were determined using a TGA-2000 thermogravimetric analyzer (TGA) (Navas Instruments, Braz.J. Food Technol., Campinas, v. 26, e2023029, 2023 | https://doi.org/10.1590/1981-6723.029235/13 Conway, USA) at 105 °C and 550 °C, respectively, up to constant weight.Lipid content (LIP) was measured using the Soxhlet method 945.38, total nitrogen was measured using the Kjeldahl method 2001.11, protein content (PRO) was estimated as 5.75 × total nitrogen, and dietary fiber (FIB) was measured using the enzymatic-gravimetric method 985.29.The Carbohydrate content (CHO) was calculated from the difference.

Experimental design and statistical analysis
The experiment was arranged in a 2 3 factorial design with four augmented central points.The established levels were chosen based on preliminary trials and consulted the literature.The independent variables (X 1 : whole-banana flour addition, X 2 : feed moisture, and X 3 : barrel temperature in the last zone) and responses (physical properties and proximal chemical composition) are presented in Table 1.A two-interaction linear model was tested according to Equation 3, to represent the behavior of each response in terms of the studied independent variables.The statistical significance of each term in the equation was determined by Analysis of Variance (ANOVA) at 5%, and 1% and regression analysis (Table 2).
where  � is the estimated response,  ̂0 is the intercept;  ̂1,  ̂2, and  ̂3 represent the linear terms;  ̂12 ,  ̂13 , and  ̂23 represent the interaction terms; and X 1 , X 2 , and X 3 are independent variables.Response surface plots were generated for the adjusted models with no significant lack of fit (LoF) and a coefficient of determination (R 2 ) > 0.70.The level curves selected for each response were superimposed to determine the optimal regions of operation from the technological and nutritional perspectives.Multiple response analysis was performed by reducing data variability to two principal components in which correlations among responses were evaluated.Statistical software (version 12.0; StatSoft, Tulsa, USA) was used to analyze the aforementioned statistical methods.Correlations below 0.5 were considered weak and above 0.8 were strong (Montgomery & Runger, 2018).

Particle-size distribution
Particle-size distributions were different for the WBF and RF (p < 0.05, Figure 1a).The WBF had 59% of particles ≤ 75 μm, whereas the RF had 31% of particles between 106 and 150 μm.The differences in particle size could be attributed to the composition of the food matrix between WBF and RF (Table 1).Endosperm hardness in rice grains is due to its starch granules embedded in a protein matrix (Chandrasekhar & Chattopadhyay, 1990), while the mesocarp in banana pulp is rich in soluble fibers (Khoozani et al., 2019), and its starch is poorly associated with protein.These structural tissues are broken in different ways during the grinding process, resulting in different shapes and particle-size distributions for both ingredients (Vargas-Solórzano et al., 2020).Particle size influences the water absorption rate.Particles smaller than 180 μm absorb more water during material wetting (Oladunmoye et al., 2014).The flours used in the experiment had 85.1% and 77.8% particles ≤ 150 μm for WBF and RF, respectively, which allowed uniform wetting before extrusion processing.2), the process variability was explained by the main effects of X 2 (87.92%, p < 0.01) and X 1 (5.30%, p < 0.05).The adjusted model had a non-significant LoF (p > 0.05) and good R 2 (0.932), which could be used for predictive purposes.The regression coefficients of the fitted models are presented in Table 2.The  ̂2 had a high negative impact (-65.31) and the  ̂1 had a low positive impact (+16.04), on SME values (Figure 1b).The SME increases either by reducing the moisture or increasing the fiber content in the feed material.This occurs because of a lower lubricating effect at low moisture levels (Mazlan et al., 2020), and insoluble fibers are hard to melt, hindering the flow of molten material and consequently increasing the torque.The SME is also affected by the particle size that structures a feed material.Fine particles are compacted to a greater extent and can increase the torque (Carvalho et al., 2010).In this study, the feed material was composed of particles smaller than 355 μm, which could have also contributed to the increasing SME values.However, the X 3 did not affect the SME in the studied interval.During extrusion cooking experiments, temperature changes of approximately 20 °C were not significant for the SME values (Altan et al., 2008;Bastos-Cardoso et al., 2007).Lowtemperature profiles in the barrel are desirable for economizing energy and processing costs.

Effect of extrusion process on HD
Flake hardness (HD) was measured as the maximum force required to compress a cylindrical bed of flakes of a fixed height.The flakes produced from the extrudates were flattened by rollers in which the air cell structures were broken, yielding HD values between 97.02 and 292.21N (Table 1).These values are higher than the HD of commercial corn flakes (64.8 ± 6.28 N) determined using the same method.Based on the ANOVA data (Table 2), process variability was explained by all effects in the following order: X 3 (36.81%,Braz.J. Food Technol., Campinas, v. 26, e2023029, 2023 | https://doi.org/10.1590/1981-6723.029238/13 p < 0.01), X 1 (23.11%,p < 0.01), and the interaction effect X 13 (10.41%,p < 0.05).The linear model with the interaction test showed a non-significant LoF (p > 0.05) and acceptable R 2 (0.711), which could be used to predict future values within the studied range.Changes in X 3 from 80 to 100 °C slightly increased HD, whereas changes in X 1 from 50 to 70% considerably decreased HD (Figure 1c).Furthermore, the interaction effect X 1 X 3 also decreased HD.The regression coefficients of the fitted models are presented in Table 2.The  ̂3 had a high positive value (+43.39) and contributed to increasing HD, whereas  ̂1 and  ̂13 showed negative values (-4.38 and -23.93, respectively) and contributed to decreasing .Extruded products based on banana/maize showed HD values from 30.2 to 56.6 N (Oduro-Yeboah et al., 2014), and51.35 to 138.39 N (Kaur et al., 2015).These low values can be attributed to the air cell structures preserved in the extruded products, unlike those in the flattened products obtained in this study.This could be attributed to the soluble fiber content present in the banana pulp, which represents 60% of the fruit's overall weight, as well as to the pectin and gums of the banana peel (Khoozani et al., 2019).It appears that the hardness of the flakes has an inverse relationship with the content of soluble fibers in the feed material.The flakes obtained from extrudates with 70% addition of whole-banana flour presented a lower HD.

Effect of extrusion process on WSI
The WSI values represent the soluble polysaccharide content released in excess water and are related to the starch conversion of extruded products (Kaur et al., 2015).When comparing the WSI of both raw materials, WBF was approximately 6.5-fold higher than that of RF (p < 0.05, Table 1).These differences may be related to the composition of sugars and soluble fibers in the bulkier fraction of both raw materials: the mesocarp in bananas and the endosperm in rice.The extrusion increased the WSI values from 4.81 to 8.62 g of soluble solids/100 g, which represented an increase from 11.86 to 79.07% when compared to raw blends (7.47 and 1.15 g soluble solids/100 g for WBF and RF, respectively, Table 1).According to the ANOVA data (Table 2), process variability was explained by the main effects of X 2 (52.6%, p < 0.01) and X 1 (35.09%,p < 0.01) (Table 2).The linear model used in the experiment showed an acceptable R 2 (0.877) and nonsignificant LoF (p > 0.05).Changes in X 1 from 50% to 70% slightly increased WSI.A similar behavior was observed during the production of extrudates based on corn and banana (Kaur et al., 2015).In contrast, changes in X 2 from 27% to 33% considerably decreased the WSI (Figure 1d).As confirmed by other studies, feed moisture has an inverse effect on the WSI of extruded products (Hagenimana et al., 2006;Sarawong et al., 2014).By increasing the moisture content of the feed material in the barrel temperature range of 80-100 °C, the starch conversion was reduced.This is because the greater lubricating effect of water prevents the breaking of starch granules and limits their gelatinization (Leonard et al., 2020).The  ̂2 had a high negative value (-0.99) and contributed to decreasing WSI, whereas  ̂1 showed a positive value (+0.81) and contributed to increasing WSI (Table 2).

Effect of extrusion process on paste viscosity
The paste viscosity curve reveals the degree of gelatinization and molecular degradation of the starch granules (Carvalho et al., 2010).PV is the maximum on the viscosity curve in the heating range of 95 °C and establishes the limit between swelling and breakdown of the starch granules (Figure 2a).The peak time of WBF (8.4 min) was slightly shorter than that of RF (9.9 min).The PV values for the extruded products ranged from 562.5 to 826.5 mPa s, which decreased by 66.2% to 77.0% with respect to the raw flours (Table 1).Extruded products based on grain and banana flours showed PV values in the range of 298.50 to 703.00 mPa.sThese differences can be attributed to the extrusion conditions.At higher feed moisture levels, fewer starch granules were converted; hence, the PV of these extruded flours was higher.In addition, as more WBF was added, the insoluble fiber content hindered the swelling of starch granules, and the PV of these extruded flours was lower (Figure 2a, b).On the other hand, the SB represents the increase in the viscosity curve during the cooling period.Differences in the SB of the raw materials showed that RF was approximately 2-fold higher than that of WBF (p < 0.05).According to the Braz.J. Food Technol., Campinas, v. 26, e2023029, 2023 | https://doi.org/10.1590/1981-6723.029239/13 domain explored to produce flakes, SB values ranged from 657.75 to 880.75 mPa s, which decreased by 76.7% to 82.6% with respect to the raw flours (Table 1).Extruded products based on grain and banana flours showed SB in the range of 437.5 to 1,421.5 mPa s (Oduro-Yeboah et al., 2014).These differences can be attributed to the degree of amylose lixiviation.The starch granules that are less converted in the extruded flour have the capacity to leach amylose, which retrogrades and raises the setback viscosity.In contrast, the insoluble fiber content in the extruded flour hindered amylose retrogradation and decreased the setback viscosity (Figure 2a, c).From the ANOVA data (Table 2), the main effects of X 2 (PV: 51.95%, SB: 68.75%) and X 1 (PV: 26.67%, SB: 21.01%) were significant (p < 0.01).The linear model used in the experiment showed acceptable R 2 (0.786 and 0.898 for PV and SB, respectively), a significant LoF for PV (p < 0.05), and a non-significant LoF for SB (p > 0.05).Changes in X 2 from 27 to 33% considerably increased PV and SB, whereas changes in X 1 from 50 to 70% slightly decreased PV and SB (Figure 2b, c).The  ̂2 had high positive values (+70.13 and +61.78 for PV and SB, respectively) and contributed to increasing both responses, whereas the  ̂1 showed low negative values (-50.25 and -34.16 for PV and SB, respectively) and contributed to decreasing both responses (Table 2).Figure 2b shows the degree of extrusion cooking according to the studied variables.Figure 2c shows the reorganization tendency of the starch polymers present in the banana-rice-based suspensions that were cooled.Extrudates produced at low levels of WBF and high levels of feed moisture maintained a greater amount of non-gelatinized or partly gelatinized starch granules, characteristic of less cooked products, associated with high PV (Sarawong et al., 2014), and high values of SB (Hagenimana et al., 2006).

Effect of extrusion process on proximal chemical composition
The proximal chemical composition of the raw materials showed that WBF had approximately 10-fold more FIB, 2.5-fold more ASH, and half PRO, compared to RF (Table 1, p < 0.05).This difference could be attributed to the presence of more minerals complexed with fibers in the WBF (Kiewlicz & Rybicka, 2020).These complexes were more noticeable in the fruit peel and cereal pericarps.After extrusion, the proximal chemical composition of the extrudates was affected only by X 1 .Compared to the starting blends, ASH and LIP contents decreased, whereas the varied inversely to X 1 levels, and the PRO remained unchanged (Figure 3).Gamlath (2008) also observed decreases in ASH and LIP contents during extrudate production based on blends of RF and banana flour at different ripening stages.ASH content of extrudates ranged from 1.89 to 2.69 g/100 g, and FIB content from 6.25 to 7.51 g/100 g (Table 1).Gamlath (2008) reported ASH values for extrudates within the range found in this study; however, the FIB values were lower and varied according to the banana ripening stage in the mixture.The changes in FIB as a result of extrusion cooking are not well understood (Leonard et al., 2020).
. The changes observed in this study (Figure 3d) can be attributed to the excess fiber content in the feeding blend (X 1 : 70%), which contributed to the formation of low-molecular-weight soluble fibers, as confirmed by the high WSI values (Figure 1d), which could not be recovered by alcohol precipitation during dietary fiber analysis.All extrudates obtained more than 5 g/100 g of dietary fiber and could be considered fiber-rich food sources (Li & Komarek, 2017).These fibers would include pectin, cellulose, lignin, and hemicellulose (Khoozani et al., 2019).Increasingly, consumers are looking for alternative gluten-free products rich in fiber to benefit their health.Fiber intake reduces the risk of cardiovascular disease, intestinal disorders, and colorectal cancer (Leonard et al., 2020).ANOVA (Table 2) confirmed that X 1 was the only factor that affected the proximal chemical composition of the extrudates (Figure 3), except for CHO, where the interaction term X 13 was also influenced (Table 2, Figure 1e).The process variability due to X 1 was 87.06% for ASH, 94.39% for PRO, 73.53% for FIB, and 73.68% for CHO (p < 0.01), whereas that for LIP was 71.43% (p < 0.05).The linear model of CHO showed an acceptable R 2 (0.803), and non-significant LoF (p > 0.05), in which the negative values of  ̂1 (-0.46) and  ̂13 (-0.13) contributed to decreasing CHO content (Table 2).Furthermore, the CHO content was slightly affected by barrel temperature.By changing the temperature from 80 °C to 100 °C, the CHO increased or decreased, depending on whether the feed material had lower or higher WBF content, respectively.

Optimization process and principal component analysis (PCA)
Optimal extrusion conditions were obtained by superimposing contour graphs of SME, WSI, and SB over the X 1 X 2 plane (Figure 4a), as well as HD, CHO, and FIB over the X 1 X 3 plane (Figure 4b) to satisfy the technological and nutritional properties of the extruded flakes.Desirability ranges were as follows: 6.3 ≤ WSI (g/100 g, dry basis) ≤ 6.8, 756 ≤ SB (mPa s) ≤ 783, 7.2 ≤ FIB (g/100 g, dry basis), and CHO (g/100 g, dry basis) ≤ 84.7. Figure 4a shows the feasible operating region from a technological standpoint, in the range of: 54.2:45.8≤ X 1 (%:%) ≤ 70:30 and 29.6 ≤ X 2 (%) ≤ 32.6.On the other hand, Figure 4b shows the feasible operating region from the nutritional standpoint, in the range of 64.9:35.1 ≤ X 1 (%:%) ≤ 70:30, and 87.0 ≤ X 3 (°C) ≤ 100.Under these conditions, the SME ranged between 216 and 257 kJ/kg which supplied the necessary energy for intermediate cooking of worked mixtures and produced flakes with HD between 160-180 N, necessary values to maintain product integrity gradually, when the flakes are soaked with milk during consumption.The flakes obtained in these regions guarantee both technological and nutritional characteristics, which are necessary for the development of novel products.
The dimensionality reduction of all responses as a set into two principal components allowed for the retention of more than 70% of the variation present in the original responses (88.67%).According to the response correlation plot (Figure 4c), the responses that presented strong positive correlations (p < 0.01) were: PV-SB (+0.85),SME-WSI (+0.89),PRO-CHO (+0.91), and FIB-ASH (+0.88).Correlations between LIP-FIB and LIP-ASH scores were intermediate (p < 0.01).Diagonally opposite responses presented strong negative correlations (p < 0.01), such as SME-PV (-0.85),SME-SB (-0.93),WSI with PV or SB (-0.93),FIB-CHO (-0.98),ASH-PRO (-0.95), ASH-CHO (-0.93), and PRO-FIB (-0.91).The smallest and largest angles formed between the two vectors belonging to different response groups (physical or chemical), were observed for LIP-WSI (the smallest) and LIP-PV (the largest), which is interpreted as moderate correlations between them (+0.69 and -0.70, respectively, p < 0.05).HD was the only response poorly represented by the principal components and did not correlate with any physical or chemical response.The projection of the treatments on the factor map (Figure 4d) confirms the small contribution Braz.J. Food Technol., Campinas, v. 26, e2023029, 2023 | https://doi.org/10.1590/1981-6723.02923 12/13 of X 3 to the data variability; that is, the change in barrel temperature from 80 to 100 °C was not significant in most of the assessed responses because factorial points 1 to 4 are close to factorial points 5 to 8. By analyzing Figure 4c and d together, it can also be confirmed which treatments presented maxima in the response surfaces for the physical properties shown in Figure 1, 2. In this way, trials 3 and 7 showed the highest SB and PV values, trials 2 and 6 the highest WSI and SME values, and trial 5 the highest HD value.However, given that chemical composition responses were only affected by X 1 , trials 4 and 8 showed the highest ASH, LIP, and FIB values (X 1 = 70%), whereas trial 5 had the highest PRO value (X 1 = 50%).A specific point within the optimal region shown in Figure 4 a, b (X 1 -X 2 -X 3 : 66% -31% -92 °C) was also projected as a predicted treatment (PR).As shown in Figure 4d, the PR point was positioned near the center-point treatment (CP) in trials 4 and 8. Intermediate values of SME, WSI, and HD, and high values of dietary fiber were achieved at this location.

Conclusion
The addition of whole banana flour (WBF) significantly modified the technological and nutritional properties of extruded flakes based on whole-banana/rice flour blends.By increasing the WBF in the feed blend, all physicochemical properties linearly increased.WBF addition showed a synergistic effect only with barrel temperature on changes in hardness and carbohydrate content.Meanwhile, feed moisture linearly affected the physical properties, except for hardness, and did not interact with other factors.In addition, an optimal region was identified using mathematical models developed from the responses.The results suggest that the WBF should be incorporated above 64.9%(dry basis), the raw blend moisture fixed between 29.6 and 32.6%, and the barrel temperature in the last zone maintained above 87 °C to produce extruded flakes with acceptable technological and nutritional characteristics: low WSI range (6.3 to 6.8 g/100 g), intermediate HD range (160 to 180 N) and a fiber content greater than 7.2 g/100 g.Overall, this work could contribute to the development of more nutritious breakfast cereals and diversify the range of existing gluten-free and highfiber products on the market for people with gluten allergies and celiac disease.On the other hand, sustainable processing is proposed to obtain and use banana flour because the peel is used, and no waste is generated.

Figure 4 .
Figure 4. Region of the optimum found by overlaying response surfaces: (A) specific mechanical energy (SME, kJ/kg), water solubility index (WSI, g/100 g), and setback viscosity (SB, mPa s); (B) hardness (HD, N), carbohydrate content (CHO, g/100 g), and dietary fiber (FIB, g/100 g).Principal component analysis (PCA): (C) physical and chemical response correlation plot; (D) projection of the treatments on the factor map. CP: center-point treatment; PR: a predicted point within the optimal region.

Table 1 .
Experimental design and responses for physical properties and proximal chemical composition of extruded flakes and raw flours.

2 Effect of extrusion process on SME SME
values are presented in Table1and ranged from 171.6 to 345.0 kJ/kg.According to the ANOVA data (Table

Table 2 .
Sum of squares (SS) of the adjusted ANOVA and regression coefficients (in coded levels) of the adjusted models for physical properties and chemical composition of extruded flakes.