Results so far
w/papers & abstracts
Assessment of fish (cod) freshness by VIS/NIR spectroscopy
F.Sigernes, K. Heia, M. Esaiassen, J.P. Wold and N.K. Sørensen
Norwegian Institute of Fisheries and Aquaculture Ltd., 9005 Tromsø, Norway.
The use of Near infrared (NIR) spectroscopy has proven promising for assessing aspects of fish quality, e.g. fat, water, protein and salt contents. Previously, transmission measurements in the wavelength range 860 to 920 nm have been conducted on samples of cod muscle. A general correlation between attenuated light and storage time of cod fillets on ice were found.
New instrumentation using diffuse reflectance now allows measurements to be done non-constructive on whole fillets. In this study, the main aim was to provide information on how fish fillet (cod) change properties in spectral response in the wavelength region 400-1100 nm as a function of time, days on ice. Since fish muscle is highly absorbing / opaque / cloudy an approach called Diffuse Reflectance also known as Transflection was used. In this mode, monochromatic light strikes the fillet perpendicular to the fillet surface. The light then penetrates the fillet and is absorbed and diffusely reflected in all directions. The diffuse back-scattered light of the fillet was then detected.
5 fresh cods were killed. One fillet per fish was selected and put on ice. The 5 fillets left was frozen for later use. First day of measurements was 24 hours after death. Each measured cycle per day took approximately 2 hours. 5 spectral samples per side of the fillets were recorded. The fillets were then put back to storage on ice. The above procedure was repeated daily over a period of 14 days.
It was easy to observe the spectral difference between dark- and white- fish muscle. The dark muscle had high absorbance in the visible compared to the white muscle. The white muscle scattered visible light better than the dark muscle. In the near infrared region the two muscle types followed the same trend. The diffusion became smaller with increasing wavelength.
The most characteristic or dominant spectral change over time was obtained in the VIS region from the skin-side tail region of the fillets, where the probes field of view covered mostly the dark muscle. A model for predicting stored time from observed spectra was constructed using Partial Least Square Regression (PLSR). The correlation coefficients between measured and predicted time were 0.95 and 0.94 for samples of white and dark muscle, respectively. The standard deviation were found to be close to 24 hrs.
5 fillets (A,B,C,D and E) have been sliced out of 5 fresh fish. First day of measurements was 24 hours after death. Each measured cycle per day took approximately 2 hours and 5 spectral samples per side of the fillets were recorded. The fillets are then put back to storage on ice. The above procedure was repeated daily over a period of 14 days.
This experiment is a pre-test on which optical methods to use in the main scheduled program later this fall-97.
Figure 1. NIRSystem 6500 optical setup
The samples are illuminated with monochromatic light through the fiber probe. The detectors, regular CCD chips, detects the reflected / transmitted light from the sample through the fiber probe. The probe itself consists of splitted fibers grouped to receive or transmit light. The grating is diameter is 10 cm and the focal length is approximately two times larger. The slits are at least 1 mm wide.
The instrument is frequently checked to evaluate performance and stability by the test listed below:
Noise tests and wavelength calibration is carried out every day with polystyrene and dydimium paddles. (See figure below).
Figure 2. Wavelength calibration with dydimium paddle.
The ratio between the detected sample intensity I and the incident intensity J is in our case named Reflectance and is represented as
where x is the wavelength in nm. Absorbance is then related to reflectance by the following equation
Absorbance(x) = A(x)=log[1/R(x)] . 
Note that the logarithm used in equation  is related to the Beer / Bougier / Lambert Law. Equation  is visualized in fig. 2a, which simply state that high absorbance means low reflectance or, low diffuse back-scattered light.
The incident intensity or the reference spectrum is obtained by holding a piece of white diffuse Teflon in front of the probe instead of the actual sample. It is taken before and after each measured cycle to make sure that the lamp has been stable during operation. The lamp is temperature stabilized. Figure 3. show a typical reference spectrum from day 3. The spectrum from the opal glass plate used between the fillet and the Teflon plate is shown in fig. 3a ( see fig.1.).
Figure 3. Typical reference spectrums.
Each stored spectrum is an average of 32 spectra. The instrument also provide auto bias and gain to obtain a maximum signal to noise ratio. A simple test was performed to evaluate how equation  turns out when we assign the sample to be the same as the reference. In other words, we test the reference quality between two measurements.
Figure 4. A simple system test.
The noise in the short-wave region of the visible is not to bad but significant larger than compared to the near infrared region. I cannot at this stage explain this effect but it may just be a default characteristic of the detector itself. The detectors are not cooled.
Results so far !
Figure 5. shows a typical data set from a fillet on day 3. The dark muscle has a high absorbance in the visible compared to the white muscle. The white muscle scatter visible light better than the dark muscle!. In the near infrared region the two muscle types follow the same trend. The diffusion becomes smaller with increasing wavelength. Also, the thickness of the fillet is related to the level of absorbance. This effect is clearly seen by the shift towards slightly higher absorpance level as we sample thicker and thicker portions of the fillets.
Figure 5. Skin-side and in-side spectra of cod-fillet
Simple eye observation: During the experiment I noticed that the back-scattered light from the in-side fillet close to the probe was kind of orange/yellow. It actually looked in color like the Mercury doublet at 577.1/579.2 nm or the Sodium doublet 589.2/589.8 nm. The light had no problem penetrating the fillet!. If I turned the fillet over to the skin-side the light turned more weak and reddish. This observation is indeed consistent with figure 5.'s minimum in-side absorbance above the blue part of the spectrum and the more red minimum at the skin-side of the fillets. PS! In a Sodium lamp the doublet lines are the only visible lines!
The most characteristic or dominant spectral change over time was obtained from the skin-side tail region of the fillets. The probes field of view covers only the dark muscle in this region. Note, that on the skin-side of the fillet the tail is darkest and as we move up it becomes more white. The inside of the fillets consists of white muscle only. Figure 6 shows spectra form the skin side tail region as a function of ice days. The color scale to the right indicate the day number. The color blue is day number one.
Figure 6. Skin side tail spectra as function of stored days for fillet (A). Click on figure to view rest of fillets.
It is quite clear that a large change appear in the visible region. To me it seems like we are loosing spectral structure with increasing time. The dip in the spectra at approximately 500 nm becomes less and less dominant. This is also seen in the peaks at day one close to 600 nm. They just vanish.
Figure 7. shows spectra from the inside and thickest part of the fillets. In this case it is hard to see any change in spectral characteristics due to time. We will therefore later use Principal Component Analysis (PCA) and Partial Least Square Regression (PLSR) to reduce any trends in time that may be hidden for the naked eye
Figure 7. Inside spectra as a function of stored days on ice for fillet (A). Click on figure to view rest of fillets.
The first week of measurements are grouped together very close. However, a small change did occur from day 11 to 14. The visible absorbance increased a little while the near infrared correspondingly decreased.
Principal Component Analysis (PCA) and Partial Least Square Regression (PLSR)
The first analysis of the thickest part on the inside of the fillets have been
carried out. A correlation between meaured and predicted time has been found. This
was expected and the results are shown
Also, on the skin tail side of the fillets we have a correlation
between predicted and measured time.
Papers and abstracts based on the above experiment