laser scanning system for real-time mapping of fiber formations in meat analogues

E:Fo

odEn

ginee

ring&

Phys

icalP

rope

rties

JFS E: Food Engineering and Physical Properties

Laser Scanning System for Real-Time Mappingof Fiber Formations in Meat AnaloguesJ. RANASINGHESAGARA, F.-H. HSIEH, H. HUFF, AND G. YAO

ABSTRACT: High moisture extrusion has been used to produce vegetable meat analogues that resemble real animalmeat and can provide significant health benefits. Since visual and textural properties are key factors for consumeracceptance, assessing fiber formation in the extruded products is important for quality control purpose. Recently,we developed a nondestructive photon migration method to quantify fiber formation in meat analogues. In thisstudy, we implemented this technique in a real-time optical scanning system. This system can scan the entire samplearea in real-time and provide 2-dimensional maps to visualize the degree of fiber formation and fiber orientationin the sample. The new system has a potential to provide a fast, nondestructive means for online monitoring of thefiber formation in meat analogues.

Keywords: diffuse reflectance, fiber formation, meat analogues, optical scanning, photon migration

Introduction

Soy protein is one of the major vegetable proteins and is abun-dantly available at relatively low cost (Liu and Hsieh 2007). In

addition, it contains no cholesterol and very low saturated fat. Thusit is an ideal protein source for people who are restricted of eat-ing meat proteins due to health or religious reasons. Under highmoisture conditions (typically 50% to 80%), a twin-screw food ex-truder with a smooth barrel can be used to convert soy proteins intomeat analog products that resemble chicken or turkey breast meat(Liu and Hsieh 2008). The key factors that affect consumer’s accep-tance of meat analogs are the products’ palatability and abundantfibrous structures. Therefore assessing the fiber formation in high-moisture extrusion products, preferably on a production line, be-comes an important issue.

Several techniques have been developed to assess fiber forma-tion in high-moisture soy meat analog products. Electron and op-tical micrographs have been obtained on dissected samples toidentify the fiber formation (Gwiazda and others 1987; Noguchi1989; Cheftel and others 1992; Lin and others 2000, 2002). Visualinspection of peeled samples is a commonly used method in labo-ratories (Roussel 1996; Akdogan 1999; Maningat and others 1999).This subjective method is destructive and slow, and thus is inappro-priate for industrial settings. The visual inspection can be reliablyreplaced by using an image processing technique based on Houghtransform (Ranasinghesagara and others 2005). However, it still re-quires preprocessing of samples before analyzing. Yao and others(2004) developed a technique based on florescence polarizationspectroscopy for detecting fiber formation in high-moisture soyprotein extrudates. This method is nondestructive and objective.However, several measurements with different polarization direc-tions are needed to derive the degree of fiber formation. In addition,the fluorescence signal is weak and the ambient light has to beblocked, which is not convenient in a production line.

MS 20080643 Submitted 8/26/2008, Accepted 10/21/2008. Authors Ranas-inghesagara, Hsieh, and Yao are with Dept. of Biological Engineering andauthors Hsieh and Huff are with Dept. of Food Science, Univ. of Mis-souri, Columbia, MO 65211, U.S.A. Direct inquiries to author Yao (E-mail:[email protected]).

Recently, we have developed a nondestructive method to de-tect fiber formation in meat analogues by measuring the opticalreflectance pattern at the sample surface (Ranasinghesagara andothers 2006). When a light is incident on an extrusion sample, it isscattered and absorbed inside the sample (Ostermeyer and Jacques1997; Tuchin 2000; Raghavachari 2001). Sample internal structuresplay a great role on modulating the light scattering process. In fi-brous samples, the light scattering probability is directional de-pendent. Specifically, when a photon is incident along the fiberdirection, its chance of being scattered away from the incident di-rection is smaller than when the photon is incident perpendicu-larly to the fibers. Photons can travel longer distances along thedirection associated with smaller scattering probabilities. Whenscattered photons reach the sample surface, they carry useful in-formation about the sample internal structure.

Continuous time random walk (CTRW) theory (Dagdug and oth-ers 2003) has been successfully applied to describe light propa-gation in fibrous anisotropic samples such as skin and collagentissues (Sviridov and others 2005). The CTRW theory determinesthat the equi-intensity distribution of the surface optical re-flectance in an anisotropic sample has an elliptical pattern (Dudkoand others 2004). The long axis of the ellipse is oriented alongthe fiber direction inside the sample. The ratio of the 2 ellipticalaxes can be related to the degree of fiber formation. In our previ-ous study on meat analogs (Ranasinghesagara and others 2006),we have shown that results obtained from this photon migrationmethod are in excellent agreement with the actual fiber formation.This technique is fast and noninvasive, thus is ideally positioned foronline monitoring of fiber formation in soy meat analogs.

In this study, we further advanced this technology by imple-menting it in a fast laser scanning system. The LED light sourceand the optical fiber described in our previous study (Ranasingh-esagara and others 2006) were replaced by a laser to increasethe incident light intensity. This new implementation allows 2-dimensional mapping of the fiber formation and orientation in theentire sample in real time. The system has been tested on extru-sion products produced from the high moisture extrusion facilityin the food engineering lab at Univ. of Missouri, Columbia, Mo.,U.S.A.

C© 2008 Institute of Food Technologists R© Vol. 74, Nr. 2, 2009—JOURNAL OF FOOD SCIENCE E39doi: 10.1111/j.1750-3841.2008.01032.xFurther reproduction without permission is prohibited

E:FoodEngineering&PhysicalProperties

Mapping of fiber formations in meat analogues . . .

Materials and Methods

MaterialsThe extruder and extrusion conditions were similar to those de-

scribed in our previous study (Yao and others 2004). A co-rotating,twin-screw food extruder (MPF 50/25, APV Baker Inc., GrandRapids, Mich., U.S.A.) with a smooth barrel and a length/diameterratio of 15:1 was used to produce extrusion samples. The barrelwas sectioned into 5 temperature zones and each zone was heatedwith an electric cartridge heating system. A long cooling die wasmounted at the end of the barrel and was cooled by a 50:50 mixtureof cold water and ethylene glycol at 5 ◦C.

We used 2 different raw mixtures in this study. The raw mate-rials used in mix nr 1 were 85% soy protein isolate (Profam 974,ADM, Decatur, Ill., U.S.A.) and 15% wheat starch (MGP Ingredients,Atchison, Kans., U.S.A.). In mix nr 2, the amount of soy protein iso-late was decreased to 57% and the wheat starch was decreased to5%. In addition, 38% wheat gluten (MGP Ingredients) was added tothe mix. All mixtures were extruded at 165 ◦C with 60% and 70%(wb) moisture contents in the aforementioned high-moisture twin-screw extruder.

Scanning systemA schematic illustration of the scanning system is shown in

Figure 1. A nonpolarized He–Ne laser (10 mW, 633 nm) was usedas the light source. The incident beam was slightly focused by a fo-cusing lens (nr 2, f = 450 mm) to provide a small beam size on thesample surface. The laser beam was redirected by a reflection mir-ror (nr 3) to a galvanometer scanner (M2, General Scanning, Biller-ica, Mass., U.S.A.) after passing through a small aperture (1 mm)at the center of the imaging mirror nr 4. The incident light wasscanned across the extrusion sample by rotating the galvanome-ter. The beam diameter of the incident laser beam is approximately1 mm at the sample surface. The diffusely reflected light from thesample was redirected by the scanning mirror and the imaging

Figure 1 --- Schematic of the scanning system. 1: laser; 2:focus lens; 3: reflection mirror; 4: mirror with a centerhole; 5: camera; 6: scanning mirror; 7: extrusion sampleout of the die.

mirror nr 4 before being captured by a CCD camera (DALSA DS-21-01-M60, Canada). A 2 × 2 binning was applied to capture a 512 ×512 12-bit image. The image was acquired by a frame grabber (So-lios XCL, Matrox Imaging, Canada) and stored in a 2-dimensionalarray in the memory for further processing. A band-pass opticalfilter at 633 nm (bandwidth = 2.4 nm, N47–494, Edmund Optics,Barrington, N.J., U.S.A.) was incorporated inside the imaging lensof the CCD camera so that the room light has little effect on theacquired images. The spatial resolution of the imaging system forboth X- and Y -axis is 96 μm/pixel. The optical system was carefullyadjusted to ensure alignment.

A coordinate system was defined so that the extruder axis wasalong the Y -axis and the extrudates moved at a constant speedalong this direction. At the beginning of each scan, the galvanome-ter was set at the boundary of the sample and then moved alongthe X-axis toward the other side (Figure 2). When reaching thesample boundary, the galvanometer was stopped and moved back-ward toward the initial position. This process was repeated untilthe system was stopped by the operator via a software interface.Considering the extudate’s linear movement, 2 different scanningmodes can be realized: the zig-zag and parallel scan. A zig-zag scanis formed when image acquisition takes place during both forwardand backward scanning so that the scanning trajectory forms a zig-zag pattern on the sample surface. A parallel scan is achieved ifimage is acquired only during the forward scan and the scanninglocations form a series of parallel lines at the sample surface. Itcan be seen from Figure 2, that a zig-zag scan provides additionalscans between 2 parallel scans although the spatial locations of thesampling locations are not evenly distributed. Therefore, the fiberformation images cannot be constructed from a zig-zag scan asstraightforward as from a parallel scan. All results described sub-sequently were obtained by using parallel scans.

The scan angle of the galvanometer is directly proportional toapplied voltage. The camera and the scanner are synchronized us-ing 2 signal sequences generated from a signal generation board(PCI-6221, Natl. Instruments, Austin, Tex., U.S.A.). The 1st signal isa TTL sequence that triggers the frame grabber. For a parallel scan-ning, the frame grabber is only triggered during the 1st half of thescan cycle as shown in Figure 2. But it is continuously triggered dur-ing a zig-zag scanning pattern. The 2nd signals produce discretevoltages to move the scanner. If the distance from scan mirror tothe sample is L, the step angle (�θ) of the scanner can be calculatedas:

�θ = tan−1( s

L

), (1)

where s is the scanning spatial resolution or the distance on thesample surface between 2 consecutive scanning points. The stepvoltage increment/decrement (�U) required to produce a �θ canbe calculated as:

�U = �θ

G, (2)

where G is the responsivity of the scanner in degree per Volts. Themaximal image acquisition speed of the system is automatically ad-justed in the control software based on specified scanning spatialresolution (s), extrusion speed (v), and scanning width (d). The to-tal number of scanning points (Nline) within a single scanning lineis calculated as:

NLine = ds, (3)

E40 JOURNAL OF FOOD SCIENCE—Vol. 74, Nr. 2, 2009

E:Fo

odEn

ginee

ring&

Phys

icalP

rope

rties


where the scanning width d is usually a little smaller than the sam-ple width to avoid any boundary effects. In a parallel scanning, thescanner is set to complete 1 full cycle when the extrusion sampleprogresses a distance similar to the scanning spatial resolution s. Inother words, the scanner has moved a total of 2 × Nline points whenthe sample has moved a distance of s. Therefore the frequency ofthe signal (fSignal) generated to drive the galvanometer is

fSignal = 2 × NLine × vs

= 2dvs2

, (4)

where v is the extrusion speed. The signal generation frequencyshould be smaller than the camera frame rate (fSignal ≤ fRate) to avoidany missing frames during the scanning. Since the small step re-sponse time of the scanner is less than 400 μs and the image pro-cessing takes approximately 3 to 5 ms, the image acquisition speedin our system is limited by the frame rate of the camera, which is60 Hz. The actual image acquisition speed used for this study was20 Hz because the extrusion speed was 10 mm/s, the sample widthwas 60 mm, and the scanning spatial resolution was 7 mm.

Calculation of the fiber formationFigure 3 shows the image and data processing sequence. When

triggered, the camera captured the optical reflectance image, whichwas stored in a temporary array. A predefined pixel intensity wasapplied to extract all image pixels with similar intensities. A ±2%intensity margin was used in the study during the pixel search tocompensate for intensity variations. The direct least-square-errorellipse fitting method (Fitzgibbon and others 1999) was then ap-plied to obtain the best-fit ellipse for those extracted pixels. A

Figure 2 --- Signal sequences forsynchronizing the optical scanner andthe CCD camera for 1 scan cycle. Aparallel scanning sequence wasillustrated, where the CCD is triggeredonly during the forward scanning. For azig-zag scanning, the CCD is triggeredduring both forward and backwardscanning. The scan patterns are alsoshown with circles representingscanning points on the sample.

Acquire reflectance image

Find equi-intensity pixels

Record their coordinates

Obtain “best fit” ellipse

Calculate B and angle

Figure 3 --- Image anddata processingprocedures.

complete set of ellipse parameters such as the center coordinates,length of axes, and the ellipse orientation angle (relative to the Y -axis) was obtained. The Bias parameter B is calculated as

B =(

L L

L S

)2

, (5)

where LL and LS are the long and short axes of the best fit ellipse.The calculated B parameter and orientation angle were displayed inreal time in the software. As shown in our previous studies (Ranas-inghesagara and others 2006), the B parameter has a good cor-relation with the degree of fiber formation. A larger B parameterindicates samples with better fibrous structures; while a B param-eter close to 1 indicates samples with no fiber formation. The longaxis of the ellipse indicates the fiber orientation. Ideally, the fiberformation is aligned with the extrusion direction (the Y -axis in Fig-ure 1). Therefore, quantitatively, the fiber orientation angle, α, isdefined as the angle between the Y -axis and the long axis of theellipse. For a sample with uniform fiber formations along the extru-sion direction, α is close to 0◦.

Two-dimensional mappingOnce the B parameters and orientation angles were obtained

for all scanning positions, these values were used to construct 2-dimensional maps. The measurement results were organized in a2-dimensional array based on the original measurement locations.A pseudo-color coding was used to display B values. In other words,a different color was used to represent a different B value. To dis-play fiber orientations, a vector map was used. A small arrow was

Incident angle (degree)

0 8-8 0

Bia

s p

ara

me

ter

0.0

0.5

1.0

1.5

2.0 Figure4 --- TheBias pa-rametersobtainedat 3differentlightincidentangles.

Vol. 74, Nr. 2, 2009—JOURNAL OF FOOD SCIENCE E41



drawn at each scanning point and the arrowhead represented thefiber orientation.

The 2 maps of the fiber formation and fiber orientation were co-registered based on the locations of the scanning point. Since thesample was moving during the scan, the calculated 2-dimensionalmaps were slanted. These 2-dimensional maps provide convenientview for assessing the fiber formation and fiber orientation in theentire scanning area.

Results and Discussion

We conducted several offline tests to verify the performance ofthe laser scanning system and studied the effects of the laser

incident angle and sample orientation. We also conducted real-time scanning experiments on fresh extruded samples and studiedthe fiber formation changes during the cooling phase.

A scan step of s = 7 mm was used in our tests. A smaller scanstep may provide a better scanning spatial resolution. On the other

Figure 5 --- The effect of sample fiber orientation on themeasured bias parameter. The top pictures showed thatextracted equi-intensity pixel data (points) and the bestfit ellipse (line) for the sample at different orientation.

Figure 6 --- The 2-dimensional maps of the B parameter and fiber orientation angle obtained in 2 extrusion samples.I, II, and III are 3 different measurements of the same sample. (A) Sample extruded with 70% moisture content. (B)Sample extruded with 60% moisture content. The mix nr 1 was used in the extrusion.

hand, this value should not be much smaller than the size of theextracted ellipse. Our previous study (Ranasinghesagara and others2006) suggested that a smaller than 5-mm scanning spatial resolu-tion may lead to an unstable B parameter, whereas a value largerthan 8 mm led to noisy data due to the low-power LED source used.When the measurement location was too close to the incident lo-cation, the results were more subject to the effects of surface ir-regularities. In this study, we improved the signal to noise ratio byusing a 10-mW He–Ne laser. And we found that a 7-mm scanningresolution can provide stable and reliable results in the samples wetested.

Effects of incident angle and fiber orientationDuring the angular scanning, the laser beam may have slightly

different incident angles at different scanning positions on thesample surface. Especially at the boundary, the sample surfaceformed a noticeable angle with the optical axis of the camera. Thisslight angle difference might induce a distortion to the acquired im-age because our imaging axis is aligned with the normal incidentlight. However, in the actual instrumentation, this angle is usuallyabout several degrees and thus its effect is negligible. To reveal anypotential effects, we imaged a fixed point at a sample extruded with60% of moisture and mix nr 2 at different incident angles. At eachincident angle, the same point was scanned 25 times. The mean andstandard deviation of the bias parameter B are shown in Figure 4.The results show no significant effect.

The fibers inside extrudates may not be aligned with the extru-sion direction. If the imaging system is not perfectly aligned, theacquired images may be spatially distorted and the extract ellipsemay depend on the fiber orientation. We confirmed our systemalignment by testing the same sample at different orientations. Atesting sample was initially placed at a random orientation. Thenit was rotated in a step of 30◦ clockwise. At each orientation, thesame location was scanned 25 times. Figure 5 indicates that the biasparameter B does not change with the sample orientation. The ex-tracted equi-intensity data and the best-fit ellipse for each orienta-tion are also shown in Figure 5. It can be noticed that the extractedpixel points are noisy due to the speckle phenomenon caused bythe coherent light source. However, when scanning moving sam-ples in real-time tests, the sample moved a small distance of 0.5mm during the image acquisition process. The speckle noise waslargely reduced because of the averaging effect (Iwai and others1982), which resulted in smooth equi-intensity profiles.


E:Fo

odEn

ginee

ring&

Phys

icalP

rope

rties


The repeatability of the measurementsTo test the repeatability of the measurements (Perez and others

2006), we scanned the same sample area multiple times. All sam-ples were kept in room temperature for several hours to stabilizebefore taking measurements. The mix nr 1 was used in the extru-sion. The 1st sample (Figure 6A) was extruded with 70% moisturecontent and the 2nd sample (Figure 6B) was extruded with 60%moisture content. Following the procedures described in Section“Two-dimensional mapping,” the 2-dimensional maps of the biasparameter and the fiber orientation were constructed. The colorscale used to encode the bias parameter B was also shown in the

Figure 7 --- The maps of difference inbias parameter (�B) and fiberorientation (�α) calculated betweenthe first 2 measurements shown inFigure 6. (A) Sample extruded with70% moisture content. (B) Sampleextruded with 60% moisturecontent.

Figure 8 --- The Bparameter andorientation maps ofa fresh extrusionsample taken at 0,5, 10, and 15 minafter coming out ofthe extruder,respectively. Thesample wasextruded using mixnr 2 at 65%moisture content.

figure along with the maps. For example, the blue color refers tolow fiber formation and the red color refers to good fiber forma-tion. The fiber orientation at each scanning location was displayedwith small arrows.

Previous studies have revealed that the fiber formation and ori-entation may change due to many extrusion parameters. The ma-jor parameters affecting fiber formation are temperature, moisture,and the specific mixture of the raw materials (Lin and others 2002).The samples extruded with high moisture tend to have a low fiberformation with random fiber orientations. These effects were alsoobserved in this study as shown in Figure 6, that is, the bias




parameter was larger in the sample extruded with 60% moisturecontent (Figure 6B) than in the sample extruded with 70% mois-ture (Figure 6A). In addition, the fiber formation and orientationswere not uniform in the extrudates. The fiber orientations in thesample with low fiber formation (Figure 6A) were noticeably morerandom than those in the sample with good fiber formations. Fur-thermore, it can be seen that the fiber orientations in Figure 6b werenot straight. The fibers in the top portion of the sample in Figure 6bshowed a clear curvature.

Figure 6 shows 3 repeated scans of the same sample. For bothsamples, the images obtained in 3 separate scans look very simi-lar to each other. To quantify the difference between multiple mea-surements, we calculated the difference of the bias parameter B andthe orientation obtained at all scanning locations in the 2 scans.To display the results obtained in the entire scanning area, thecalculated images of �B and � α were shown as 2-dimensionalimages in Figure 7. In other words, the value of a point in thedifference image (Figure 7) indicates the difference between thevalues in the 2 original images (Figure 6) at the same point. Apseudo-color coding was used to display the numerical values of�B and � α. In the color scale used (Figure 7), green color indicatessmall or no difference between scans, red color represents a largepositive difference, and blue color represents a large negative differ-ence. The maximum changes of the bias parameter (�B) were 0.109and 0.045 in samples extruded with 60% and 70% moisture content,respectively. The corresponding average values of the coefficientof variance (CV) were 1.55% and 2.98%, respectively. The changein orientation was larger than in the bias parameter. The major-

Figure 9 --- The �B and �α map at (A) 5, (B) 10, and (C) 15 min measured relatively to the 0-min results in Figure 8.The corresponding mean value and standard deviation of the difference obtained from all image pixels were shownbelow each image.

ity of orientation angle changes were within ±15◦ (green color) inboth samples. The average values of the CV were 9.76% and 5.71%in samples extruded with 60% and 70% moisture content, respec-tively. However, a careful analysis indicated that the locations withhigher �α values had very low fiber formation. The extracted el-lipse was close to a circle when fiber formation was poor, whichresulted in an unstable orientation angle. The aforementioned test-ing results indicated that the laser scanning system showed a goodrepeatability.

Simulated conveying belt scanningA simulated conveying belt was constructed using a step-motor

driven translational stage to measure the same sample multipletimes. The speed of the stage was adjusted to match that of the ex-truder (10 mm/s). The stepper motor was programmed to bring thestage back to its original position after each scan. The movementaccuracy of the translation stage was verified by scanning a calibra-tion grid. The sample was kept on the stage throughout the entiremeasurement period so that the same points were scanned eachtime. The laser scanning system (Figure 1) was mounted above thetranslational stage. The raw material mix nr 2 was extruded at 65%moisture content. The extrudates were taken immediately from theextruder and placed on the stage that was moving at the same speedof the extruder. The entire sample was then scanned every 5 min inthe first 15 min. Figure 8 shows the actual sample and the scanningresults at 0, 5, 10, and 15 min after the extrusion. Similar to the re-sults in Figure 6, the blue color shows low fiber formation and thered color shows good fiber formation.


E:Fo

odEn

ginee

ring&

Phys

icalP

rope

rties


Due to the decrease in sample temperature and the structuralsettling, the measured fiber structures showed quite many changesat several locations during the 15 min. To better display the vari-ations, we calculated the changes of the bias parameter and theorientation for the samples measured at 5, 10, and 15 min relativeto the 0-min result (Figure 9). Despite a few locations, the majorityof the sample areas had minor changes in the B parameter (greencolor). As an overall measure of the B parameter change, we calcu-lated the mean value of �B for all pixels. Because the �B can benegative or positive, we used the absolute value |�B| in the calcu-lation. As shown in Figure 9, the calculated 〈|�B|〉 had the largestchange in the first 5 min, and the changes were slowing down in the10- and 15-min results. On the other hand, we observed high vari-ations in fiber orientation angles. The highest variance of �α wasobserved in first 5 min. The change of fiber orientation was muchsmaller thereafter as shown in Figure 9. These changes may be re-lated to internal changes of the sample during the cooling period.

Conclusions

We developed a real-time laser scanning system based on pho-ton migration to map fiber formations in meat analogues.

The bias parameter and fiber orientation angle were obtained ateach scanning location at the sample surface and 2-dimensionalimages were formed to visually display the results. The presentscanning speed of our system is limited by the CCD frame rate of60 Hz and it can be further improved by using a high-speed, highsensitivity camera. We have applied the system to map fiber struc-ture changes in extrusion samples. We observed large fiber orienta-tion changes during the cooling phase after extrusion. These testssuggested that the system has potential to be used in time sensi-tive measurements. This nondestructive real-time scanning systemmay provide a useful means to monitor fiber formations in an in-dustrial setting.

AcknowledgmentThe project was supported by the Natl. Research Initiative of theUSDA Cooperative State Research, Education, and Extension Ser-vice, grant number 2005-35503-15401.

ReferencesAkdogan H. 1999. High moisture food extrusion. Int J Food Sci & Tech 34:195–207.Cheftel JC, Kitigawa M, Queguiner C. 1992. New protein texturizing process by extru-

sion cooking at high moisture levels. Food Rev Int 8(2):235–75.Dagdug L, Weiss GH, Gandjbakhche AH. 2003. Effects of anisotropic optical properties

on photon migration in structural tissues. Phys Med Biol 48:1361–70.Dudko OK, Weiss GH, Chernomordik V, Gandjbakhche AH. 2004. Photon migra-

tion in turbid media with anisotropic optical properties. Phys Med Biol 49:3979–89.

Fitzgibbon A, Pilu M, Fisher RB. 1999. Direct least square fitting of ellipses. IEEE TransPatt Anal Mach Intell 21(5):476–80.

Gwiazda S, Noguchi A, Saio K. 1987. Microstructural studies of textured vegetable pro-tein products: effects of oil addition and transformation of raw materials in varioussections of a twin-screw extruder. Food Microstr 6:57–61.

Iwai T, Takai N, Asakura T. 1982. Dynamic statistical properties of laser speckle pro-duced by a moving diffuse object under illumination of a Gaussian beam. J Opt SocAm 72(4):460–7.

Lin S, Huff HE, Hsieh F. 2000. Texture and chemical characteristics of soy protein meatanalog extruded at high moisture. J Food Sci 65(2):264–9.

Lin S, Huff HE, Hsieh F. 2002. Extrusion process parameters, sensory characteristics,and structural properties of a high moisture soy protein meat analog. J Food Sci67(3):1066–72.

Liu KS, Hsieh F. 2007. Protein–protein interactions in high moisture-extruded meatanalogs and heat-induced soy protein gels. J Am Oil Chem Soc 84(8):741–8.

Liu KS, Hsieh F. 2008. Protein–protein interactions during high moisture extrusionfor fibrous meat analogues and comparison of protein solubility methods usingdifferent solvent systems. J Agric Food Chem 56(8):2681–7.

Maningat CC, DeMeritt GK Jr., Chinnaswamy R, Bassi SD. 1999. Properties and appli-cations of texturized wheat gluten. Cereal Foods World 44(9):650–5.

Noguchi A. 1989. Extrusion cooking of high moisture protein products. In: Mercier C,Linko P, Harper JM, editors. Extrusion cooking. St. Paul: American Assn. of CerealChemists Inc. 343–70 p.

Ostermeyer MR, Jacques SL. 1997. Perturbation theory for diffuse light transport incomplex biological tissues. J Opt Soc Am A 14(1):255–61.

Perez JM, Schreiner S, Gorton GE. 2006. Evaluation of the VITUS smart laser scannerfor accuracy, resolution, and repeatability for clinical assessment of Pectus defor-mities and scoliosis. Proc IEEE 32nd Annu Bioeng 33–4.

Raghavachari R. 2001. Near-infrared application in biotechnology. New York: MarcelDekker. 382 p.

Ranasinghesagara J, Hsieh F, Yao G. 2005. An image processing method for quanti-fying fiber formation in meat analogs under high moisture extrusion. J Food Sci70(8):E450–4.

Ranasinghesagara J, Hsieh F, Yao G. 2006. A photon migration method for character-izing fiber formation in meat analogs. J Food Sci 71(5):E227–31.

Roussel L. 1996. Making meat products using extrusion technology. Extrusion Comm9(6):16–8.

Sviridov A, Chernomordik V, Hassan M,Russo A, Eidsath A, Smith P, GandjbakhcheAH. 2005. Intensity profiles of linearly polarized light backscattered from skin andtissue-like phantoms. J Biomed Opt 10(1):014012:1–9.

Tuchin V. 2000. Tissue optics: light scattering methods and instruments for medicaldiagnosis. Bellingham: SPIE Press. 378 p.

Yao G, Liu KS, Hsieh F. 2004. A new method for characterizing fiber formation in meatanalogs during high-moisture extrusion. J Food Sci 69(7):E303–7.


laser scanning system for real-time mapping of fiber formations in meat analogues

Documents