**1. Introduction**

Optical coherence tomography (OCT) [1–4] is becoming as a new research field to find new data to study the blood oxygenation from said vessels within inner and outer retina. Noninvasive spectrophotometric retinal oximetry has been a target for over five decades ago. Our work and results reported in this chapter are inspired from a previous experiences reported in Ref. [3]. We invite lectures to study previous reports about OCT image processing [3–8] to follow the evolution of noninvasive spectrophotometric retinal oximetry efforts. However, image processing from optical coherence tomography (OCT) requires new alternatives to analyze the absorption difference between oxyhemoglobin (HbO2) and deoxyhemoglobin (Hb) in the retinal vessels.

OCT images show an internal view of the eye. The inside view of the ocular eye is recorded through the pupil, and this let to get an image that shows the vascular system inside the ocular glove [1]. The study is performed by projecting a beam of light into the pupil to facilitate the visualization of the fundus. Fundus images let to know if patients have different eye condition, like retinal detachment, retinal thrombosis, macular degeneration, diabetic retinopathy and glaucoma. However, diagnostic steel depends on the physician experience and not in a set of new parameters that can be extracted from the OCT image that converted in new data that let to know from image gray levels to measure the oxygen saturation in the retina vein system [3, 5, 7]. The oxygen saturation that is intrinsically in the gray levels of the OCT image depends directly on the frequency that the image is obtained, which comprises a range between 570 and 600 nm [9–13]. This relationship is clearly shown in **Figures 1** and **2**. **Figure 2** shows the absorption spectrum of Hb and HbO2 [9 at 532 nm].

We have now new ideas about vascular retina oximetry (VRO). VRO can be understood as the study of oxygen saturation measurement from OCT conventional images if they are processed properly. We defined the VRO approach as a new method to measure the oxyhemoglobin (HbO2) and deoxyhemoglobin (Hb) from light reflection at different wavelengths. Light reflection from HbO2 and Hb is the gray light levels seen from the OCT images. These gray levels are not properly recognized by humans; however, OCT image processing techniques, as discussed in this chapter, are excellent tools converting conventional OCT images in new ones according to light reflectance laws. The calculus of percentage of HbO2 from OCT images is a big challenge. We present in this chapter a new method to evaluate oxygen saturation at the retina vein system.

The Eyeball structure visualized from a fundus image as shown in **Figure 1**. **Figure 1** shows the main central eye arterial just entering at the optic nerve disk. The main artery divides into the central retina artery and ciliary arteries to supply the inner and outer retina at the back of the eye. However, this image, as shown, is far away to give us information about the oxygen consumption in the vein circulatory

**Figure 1.** *Absorption spectrum of haemoglobin (Hb) and oxyhemoglobin (HbO2) [8].*

*A New Method to Manipulate Conventional OCT Images to Measure Changes in the Relative… DOI: http://dx.doi.org/10.5772/intechopen.110884*

**Figure 2.** *Plot of the absorption spectrum of Hb and HbO2 [8], with the absorption spectrum indicator at 532 nm.*

system. The human retina is a highly oxygen consuming structure. It happens that the human retina consumes oxygen faster than the brain; consequently, the retina metabolic processes are highly dependent on oxygen saturation SATO2. The choroid has the responsibility to supply oxygen to the cone and rods at the retina complex system.

The transparent structure of the eye fundus offers an extraordinary chance to extract information from the visible vein system about the oxygen going in there. The image spectrometry lets to analyze the oxygen content inside the arterial system if the image is filtered adequately to extract the quantities of oxygen going inside the haemoglobin. This is not an easy task; it is a steel and noncompletely solved problem. The key point is how to measure the absorption of light of haemoglobin and deoxyhemoglobin and from OCT images which vein colors depend on the quantity of oxygen they have in there. The spectral analysis can show both structures as filters that reflect different amounts of light in specific wavelengths depending on its chemical composition; this means more or less oxygen they transport. **Table 1** shows the different wavelengths of the laser bands required to distinguished with higher accuracy between haemoglobin (613 nm) and deoxyhemoglobin (486 nm) [14].

We show in **Figure 3** the relationship between light absorption in Hb and HbO2 and its relation with the frequency of the light beam. As said above, light absorption and consequently oxygen saturation is a function that depends on the wavelength of the beam light. The optimum frequency to find differences between Hb and HbO2 are 570 and 600 nm. Jóna Valger›ur Kristjánsdóttir describes important results of light absorption capacity in [9]. Our group and other researchers [3, 5–7, 9, 15] are interested in applying the light


**Table 1.**

*Table with the different wavelengths of the laser bands [14].*

#### **Figure 3.**

*A. Original fundus image with retinitis pigmentosa obtained at* 532*nm. B. Image resulting from applying the pseudocolor method to a fundus image with retinitis pigmentosa obtained a* 532*nm.*

absorption blood properties to find differences between Hb and HbO2 [7]. The main idea goes converting those properties in new colored OCT images where colors let the physician to see the oxygen saturation performance in the retinal vascular system [3].

### **2. Methodology**

The oxygen saturation SATO2 inside the eye circulatory system is highly correlated with the light absorption [16] in each point of the OCT images. Then, the oxygen at the HBO2 at maximum saturation levels at the corresponding wavelengths shown in **Figure 3** will show the maximum absorbance of light. This absorbance is shown in an OCT image as the light intensity output filtered by the veins. The OCT emitter light or input light to the veins has a specific frequency according to spectral answer as shown in **Figure 3**. Obviously, the Hb has the lower SATO2 since Hb is the blood returning after delivering the oxygen to the retina. Hb will absorb light according to the blue curve as shown in **Figure 3**. It is well known that arterial blood is red and venues blood is blue. This color differences and oxygen saturation can be seen too in a black and white OCT image. Now consider the image in **Figure 1** as a black and white image. Notice that the black and white image shown in **Figure 4** is now the negative version of the black and white image in **Figure 1**. Then, the light reflected from the Hb system will show the less reflected light in comparison with the light reflected by the HBO2. This means that in the black and white version of **Figure 1**, the highest white illumination will correspond to the highest SATO2, while the darkness points in **Figure 4** will show the points with the lower oxygen saturation SATO2. This difference is clearly observed in the blue and green wavelengths graphs in **Figure 3**. Then, HbO2 has the highest absorption points and are at 418, 542, 577, and 925 nm. However, the Hb has highest absorption points located at: 430, 550, 758, and 910 nm.

We have a big challenge now. This means to implement computational procedures that let estimating the SATO2 from OCT images. The estimation requires calculating true values of oxygen saturation in the eye circulatory system.

#### **2.1 Oxygen calculus of the blood**

The SATO2 in HbO2 can be calculated as the relationship between HbO2 and the HbO2 + Hb [17]. This easy relation will give us the percentage of HbO2 compared with the whole oxygen in both HbO2 + Hb as shown in Eq. (1):

*A New Method to Manipulate Conventional OCT Images to Measure Changes in the Relative… DOI: http://dx.doi.org/10.5772/intechopen.110884*

**Figure 4.** *Plot of grayscale values of Figure 3A.*

$$\text{SatO}\_2 = \frac{HbO\_2}{Hb + HbO\_2} \times 100\text{\%} \tag{1}$$

where HbO2 is the oxygen combined with haemoglobin and Hb is the deoxygenated haemoglobin. The SatO2 is multiplied by 100 to express the SatO2 in percentage. The SatO2 of blood goes to 97,5–75% at 100 mmHg for the mixed venous blood.

OCT black and white images as shown in **Figure 4** are normalized to get a set of levels going from 0 to 255. The 0 represents the highest level of darkness or lowest level of illumination. The 255 levels represent the maximum white intensity. The HbO2 with the highest SatO2 will be represented with the 255 gray level. Then, we have a new easy way to evaluate the SatO2 in the circulatory ocular globe system (COGS). This new approach considers as a vector only the pixels in the veins shown in the black and white OCT. The veins are filtered by a segmentation algorithm that converts the circulatory eye system as a numerical matrix. This matrix depending on their values between o and 255 let identify the differences in SatO2 in veins and consequently will implement new alternatives to show the physician the performance of the circulatory eye system in normal and illness conditions:

$$\text{SatO}\_2 = \frac{\sum\_{n=96}^{N-1} \frac{P\_{n\_{\text{total}}}}{255}}{\sum\_{n=1}^{N-1} \frac{p\_k}{95} + \sum\_{n=96}^{N-1} \frac{P\_{n\_{\text{total}}}}{255}} \times 100\text{\%} \tag{2}$$

where *Pntotal* is

$$P\_{n\_{\text{total}}} = \sum\_{n=96}^{N-1} \frac{P\_{n\_1}}{199} + \sum\_{n=200}^{N-1} \frac{P\_{n\_2}}{255} \tag{3}$$

Eq. 2 let calculate SATO2 from an OCT gray levels image analogously as Eq. (1) does. The OCT image converted as a vector is a matrix, where each number in there represents the proportion of light absorbance by HBO2; consequently, it is a measure of the oxygen in an specific pixel depending on its value, the 255 value, and the highest gray level has the highest oxygen absorption in that pixel and so on for lower values in the 1 to 255 gray levels scale.

*Pk* in Eq. (2) is the value of the vector in pixel k. In Eq. 2, we consider that the vector *Pk* belonging to venous blood pixels going from 1 to 95.

Pn total in eq. (3) with two variables *Pn*<sup>1</sup> and *Pn*<sup>2</sup> are the pixel values of each vector to be analyzed. In this case, *Pn*<sup>1</sup> takes the values that are delimited by the n and the dividend, and these values will vary between 96 and 199, which are the values proposed in this work for the representation of venous haemoglobin. On the other hand, *Pn*<sup>2</sup> will take the values from 200 to 255 which are the values proposed to identify the arterial haemoglobin or oxyhemoglobin. Thus, when the substitution of the variables is made, the result is *Pntotal* which is the whole amount of gray levels in the vector. This includes all the pixels representing the arterial and venous system. The arterial and venous systems are normalized to their maximum values. Hb has values from 1 to 95, HbO2 venous goes from 96 to 199, and HbO2 arterial goes from 200 to 255.

We finally arrive to a final SatO2 equation as shown below:

$$\text{SatO}\_2 = \frac{\sum\_{n=96}^{N-1} \frac{P\_{n\_1}}{199} + \sum\_{n=200}^{N-1} \frac{P\_{n\_2}}{255}}{\sum\_{n=1}^{N-1} \frac{P\_k}{95} + \left(\sum\_{n=96}^{N-1} \frac{P\_{n\_1}}{199} + \sum\_{n=200}^{N-1} \frac{P\_{n\_2}}{255}\right)} \times 100\text{\%} \tag{4}$$

As a matter of fact of Eq. 4 with the following values, *Pn*<sup>1</sup> ¼ 138, *Pn*<sup>2</sup> ¼ 225, and *Pk* ¼ 80shows SAtO2 calculation for pixel in *k* = 80:

$$\text{SatO}\_2 = \frac{^{138}\text{\textdegree}\_{199} + ^{225}\text{\textdegree}\_{255}}{^{80}\text{\textdegree}\_{95} + \left(^{138}\text{\textdegree}\_{199} + ^{225}\text{\textdegree}\_{255}\right)} \times \mathbf{100}\text{\textdegree}\tag{5}$$

$$\text{SatO}\_2 = \frac{1.5758}{2.4179} \times 100\% \tag{6}$$

$$\text{SatO}\_2 = 0.6517 \times 100\% = 65.17\% \tag{7}$$

Then in (7), 65.17% is the SATO2 for pixel in *k = 80*.

The assignment of colors to the different gray level values was as follows: 0–black, 1–32 equals purple, 33–64 violet, 65–95 blue, 96–127 cyan, 128–159 green, 160–191 yellow, 192–223 orange, and 224–255 red. Then Eq. (3) can evaluate SatO2 by regions or specific borders inside the OCT image.

We can see from the above discussion that we have a new numerical method to calculate SatO2. This calculus let describes all the pixels at the circulatory eye system as numbers that represents each one the SatO2 by pixel. The SatO2 can be expressed too as a color vein map for an easy recognition of the physician about the state of the circulatory system and its relation with oxygen performance.

### **3. Results**

**Figure 3** shows the result of applying the pseudocolor method to an image that was obtained at 532*nm* with the OPTOS California retinograph (this retinograph works with different wavelengths as shown in **Table 1**), according to gray levels shown in **Figure 4**.

#### *A New Method to Manipulate Conventional OCT Images to Measure Changes in the Relative… DOI: http://dx.doi.org/10.5772/intechopen.110884*

Notice that in **Figure 2** at 532 nm, there is no difference between the absorption spectrum of Hb and HbO2. However, our results show that differences in gray levels (128–255) let find SatO2 differences, shown as red and green SatO2 levels in **Figure 3**. This is due to the good pseudocolor sensitivity segmentation method discussed above. The same can be said with the red and blue (HbO2 and Hb) intersections points in **Figure 2**.

We find in **Figure 4** the OCT image vector shown as the blue curve. The blue vector has seven changes we recognize as slopes, and each slope was associated with the SatO2 variation in a specific region of set of gray levels [18] according to the absorbance properties of Hb and HbO2, as shown in **Figure 2**.

We show in **Figures 5**–**8** the SatO2 absorbance results from images obtained with wavelengths: 486*nm*, 635*nm*, 802*nm*, and the combination of the wavelengths 532*nm* and 635*nm*.

We show in **Figures 5**–**8** the results obtained with the pseudocolor method. Color levels show differences in SatO2 in the ocular microvasculature system. This method as shown can be applied to both fundus images and angiographies. It is important to say that **Figure 8** shows the SatO2 of the choroid, making a novel method to analyze the vasculature of the choroid from the vector obtained by the pseudocolor method.

The pseudocolor method allows to differentiate clearly the microvasculature of the fundus and then to analyze the vasculature performance from new tools.

#### **Figure 5.**

*A. Original healthy fundus image obtained at* 635*nm. B. Image resulting from applying the pseudocolor method to a healthy fundus image obtained at* 635*nm.*

#### **Figure 6.**

*A. Original ocular fundus image with ocular ischemia obtained at* 486*nm. B. Image resulting from applying the pseudocolor method to a fundus image with ocular ischemia obtained a* 486*nm.*

#### **Figure 7.**

*A. Original fundus image with retinitis pigmentosa with the combination of wavelengths.* 532*nm and* 635*nm. B. Image resulting from applying the pseudocolor method to a fundus image with the combination of wavelengths.* 532*nm and* 635*nm.*

#### **Figure 8.**

*A. Original fundus image of the choroid obtained with a wavelength of* 802*nm. B. Image resulting from applying the pseudocolor method to a fundus image of the choroid obtained with a wavelength at* 802*nm.*

We show in **Figure 9A**, with author permission, the SatO2 reported by Kristjánsdóttir et al. [4]. As a matter of comparison, we show the same image obtained at 570 nm (**Figure 9B**) and 600 nm (**Figure 9C**) applying our method with results shown in **Figures 10** and **11**.

The resulting images in **Figures 10** and **11** obtained with the method presented in this chapter show notable differences in the color assignment among Kristjánsdóttir et al. and ours; as mentioned in the methodology, the oxygen saturation formula proposed in here was applied in this assignment, and this explains the difference between results reported in [4] and ours; however, color standardization requires deeper work. It is clear that wavelength has an important role to find best results when looking HbO2 or Hb. **Figure 11** shows results with 570 nm image, and **Figure 10** shows results with 600 nm.

**Figure 12A**–**C** shows results published by Geirsdottir et al. [3], which is similar to the one obtained by Kristjánsdóttir et al. [4].

The algorithm developed in this work was applied in **Figure 12A** to see how our method compares with Geirsdottir et al. and our results are shown in **Figure 13**.

We show in **Figure 14** the pseudocolor method applied in the original **Figure 12B**.

*A New Method to Manipulate Conventional OCT Images to Measure Changes in the Relative… DOI: http://dx.doi.org/10.5772/intechopen.110884*

**Figure 9.** *Oximetry of the eye obtained at 570 nm (Figure 9B), and 600 nm (Figure 9C).*

The first results obtained by Davila-Iniesta et al. [3] will be compared with the pseudocolor method as shown in **Figures 15**–**18**. We find an improvement in the segmentation procedure. We find that pseudocolor method provides a better

**Figure 10.** *False color method applied to Figure 9C.*

**Figure 11.** *False color method applied to Figure 9B.*

**Figure 12.** *Eye oximetry obtained at 570 nm (Figure 12A), and 600 nm (Figure 12B).*

resolution with vein network and gives a whole scheme about the distribution of the SatO2 at the vein system.

**Figure 16B** shows a healthy eye, and using the pseudocolor proposed method we noticed an improvement in the vein network sight in relation to the original fundus image [3, 4]. We show too that it is possible to see the SatO2 levels converted as a color set. This will help the physician to a better understanding of the performance of the vasculature system.

**Figure 17A** shows the fundus image with macular edema. After applying the false color method, the resulting image shown in **Figure 17B** was obtained. We find that the relationship between macular edema and SatO2 plays an important role. However, this statement requires statistical analysis from enough of this illness OCT images, and statically analysis can be applied since we have converted every images in a numerical vector.

**Figure 18B** shows the fundus image with macular edema after applying the pseudocolor method. The SatO2 ratio in this image is seen to be clearer than the previous image. Also, the microvascular system and how it is affected due to this disease can be seen.

*A New Method to Manipulate Conventional OCT Images to Measure Changes in the Relative… DOI: http://dx.doi.org/10.5772/intechopen.110884*

**Figure 13.** *Pseudocolor method applied to Figure 12A.*

**Figure 14.** *False color method applied to Figure 12B.*

**Figure 15.**

*A. Fundus image of healthy eye. B. False color method applied to the fundus image [3].*

We show in **Table 2** how the eye circulatory system transformed as vector can be analyzed from excel tools an even by more advanced ones like Jupiter Notebook tools.

We show in **Figure 19** an important result. The vector means that the OCT image at different frequencies is now represented as a distribution function. We find at the X

#### **Figure 16.**

*A. Fundus image of healthy eye. B. Pseudocolor method applied to Figure 15A.*

**Figure 17.** *A. Fundus image with macular edema. B. False color method applied to Figure 17A.*

**Figure 18.** *A. Fundus image with macular edema. B. Pseudocolor method applied to Figure 17A.*

axis the 0 to 255 gray levels and at the Y axis the count of pixels belonging to that level. This new representation of SatO2 gives an exact picture about the performance of the vein eye system. Taking a look at **Figure 19** Healthy eye-1, we notice clearly the distribution of SatO2 in the fondues vein system. The maximum SatO2 values are in


*A New Method to Manipulate Conventional OCT Images to Measure Changes in the Relative… DOI: http://dx.doi.org/10.5772/intechopen.110884*

> **Table2.**

 *OCT image.*

**Figure 19.**

*Histogram of different OCT images. Horizontal axis represent the 0 to 255 gray levels and vertical axis represent the count per level.*

the range between 190 and 240 gray levels, just inside the HBO2. The lowest SatO2 levels are between 0 and 95 gray levels, belonging to the HB. **Figure 19** shows that we can find a set of clearly different SatO2 distributions, and this is an important result since we can find detailed differences numerically represented that cannot be distinguished by traditional human visual OCT analysis. This of course is not a conclusive statement; however, first results go to that direction.
