Limits...
Effects of age and blood pressure on the retinal arterial wall, analyzed using adaptive optics scanning laser ophthalmoscopy.

Arichika S, Uji A, Ooto S, Muraoka Y, Yoshimura N - Sci Rep (2015)

Bottom Line: WLR showed a strong correlation with age (r = 0.68, P < 0.0001), while outer diameter and inner diameter did not show significant correlation with age in the normal group (r = 0.13, P = 0.36 and r = -0.12, P =  .41, respectively).In conclusion, AOSLO provided noninvasive and reproducible arterial measurements.WLR is an early marker of morphological changes in the retinal arteries due to age and blood pressure.

View Article: PubMed Central - PubMed

Affiliation: The Department of Ophthalmology and Visual Sciences, Kyoto University Graduate School of Medicine, Kyoto 606-8507, Japan.

ABSTRACT
The wall-to-lumen ratio (WLR) of the vasculature is a promising early marker of retinal microvascular changes. Recently, adaptive optics scanning laser ophthalmoscopy (AOSLO) enabled direct and noninvasive visualization of the arterial wall. Using AOSLO, we analyzed the correlation between age and WLR in 51 normal subjects. In addition, correlations between blood pressure and WLR were analyzed in 73 subjects (51 normal subjects and 22 hypertensive patients). WLR showed a strong correlation with age (r = 0.68, P < 0.0001), while outer diameter and inner diameter did not show significant correlation with age in the normal group (r = 0.13, P = 0.36 and r = -0.12, P =  .41, respectively). In the normal and hypertensive groups, WLR showed a strong correlation with systolic and diastolic blood pressure (r = 0.60, P < 0.0001 and r = 0.65, P < 0.0001, respectively). In conclusion, AOSLO provided noninvasive and reproducible arterial measurements. WLR is an early marker of morphological changes in the retinal arteries due to age and blood pressure.

No MeSH data available.


Calculation of the mean vascular measurements correcting for cardiac pulsation.In order to minimize the influence of cardiac pulsation on vascular measurements, the cardiac cycle was synchronized to the adaptive optics (AO) videos using pulsation data obtained through a pulse oximeter attached to the subjects’ earlobes. (a) The cardiac cycle was divided into 5 segments (0–0.2, 0.2–0.4, 0.4–0.6, 0.6–0.8, and 0.8–1.0), and each video frame was assigned to the corresponding segment. (b) The images on the same relative cardiac cycle were extracted from the AO videos. (c) The images of each corresponding segment were averaged, and the vascular caliber measurements were obtained for every 5 segments. Finally, the vascular measurements of 5 segments were averaged and used for analyses.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4507481&req=5

f5: Calculation of the mean vascular measurements correcting for cardiac pulsation.In order to minimize the influence of cardiac pulsation on vascular measurements, the cardiac cycle was synchronized to the adaptive optics (AO) videos using pulsation data obtained through a pulse oximeter attached to the subjects’ earlobes. (a) The cardiac cycle was divided into 5 segments (0–0.2, 0.2–0.4, 0.4–0.6, 0.6–0.8, and 0.8–1.0), and each video frame was assigned to the corresponding segment. (b) The images on the same relative cardiac cycle were extracted from the AO videos. (c) The images of each corresponding segment were averaged, and the vascular caliber measurements were obtained for every 5 segments. Finally, the vascular measurements of 5 segments were averaged and used for analyses.

Mentions: Vascular caliber measurement was performed in zone B28, 0.5–1.0 disc diameters away from the optic disc margin using custom software known as ARIA, which was developed by Canon Inc., Tokyo, Japan. ARIA semi-automatically segments the retinal arterial wall borders in the AOSLO video. Our segmentation method consisted of preprocessing, rough central axis setting, precise central axis and vascular wall border detection, and measurement of the thickness of the vascular wall. Preprocessing consisted of the following steps. Pixel intensity values in the AOSLO video were transformed logarithmically to increase contrast in a vascular wall and decrease it in a nerve fiber layer. The AOSLO video was then stabilized using the inverse scan-line warping method. In order to obtain a high-contrast vascular wall image, frames of the stabilized AOSLO video corresponding to the specific phase range of the pulse wave were selected, and these were then smoothed using a 3-D median filter. We used a filter size of 3 × 3 × (selected frame numbers). In the next step, we manually set seed points at the center of the retinal artery along its running direction (Supplementary Figure S1A). At each seed point, the line segment that was perpendicular to the running direction of the seed points was set, and its pixel intensity profile was calculated. In the precise central axis and vascular wall border detection step, we applied a sliding linear regression filter (SLRF)3334 to the pixel intensity profile along the line segment (Supplementary Figure S1B). The SLRF method is based on the fitting of a line by linear regression that relates the pixel intensity value to the distance along the profile within a sliding window (window size: W). The precise position of the vascular central axis was identified around the seed point as a zero-cross point in the pixel intensity profile filtered by SLRF. To determine the vascular edge position, we specified the minimum point in the left side and the maximum point in the right side from the zero-cross point corresponding to the central axis. In order to determine the vascular wall border position more robustly and precisely, we applied SLRF twice to the pixel intensity profile along the line segment that was perpendicular to the central axis. Specifically, SLRF with a larger window size (W = 10 pixels) was applied first, followed by SLRF with a smaller window size (W = 4 pixels). The initial, larger window size (W = 10 pixels) was applied to establish the approximate position of the vascular wall region. In the pixel intensity profile filtered by SLRF, we could establish the approximate position of the vascular wall region by detecting the minimum point in the left side and the maximum point in the right side from the zero-cross point corresponding to the central axis (the minimum and maximum points are shown in Supplementary Figure S2C). Subsequently, we deleted outliers in the first candidate points if their Euclid distance from the central axis was less than the lower threshold (Tl)1 percentile or more than the higher threshold (Th)1percentile (Supplementary Figure S1C). In this study, Tl1 and Th1 were set as 10 and 90, respectively. The remaining first candidate points were interpolated using the natural spline interpolation method (Supplementary Figure S1D). The interpolated candidate points indicated the approximate position of the vascular wall region and were used to calculate the position of the vascular wall border by applying it to the second step (W = 4 pixels) (Supplementary Figure S1E). In the pixel intensity profile filtered by SLRF, we select the 2 nearest extremal points to the candidate point as vascular wall border candidate points (inner and outer borders are shown in Supplementary Figure S2D). Subsequently, we deleted outliers in the vascular wall border candidate points if their pixel intensity values were less than the Tl2 percentile or more than the Th2 percentile, and then deleted the remaining vascular wall border candidate points if their pixel intensity values were less than the Tl3 percentile or more than the Th3 percentile (Supplementary Figure S1F). In this study, Tl2 and Tl3 were set as 10, and Th2 and Th3 were set as 90. The remaining vascular wall border candidate points were sampled at intervals, and then interpolated using the natural spline interpolation method (Supplementary Figure S1G). Finally, we used the interpolated candidate points to define the vascular wall border in calculating the retinal arterial wall thickness (Supplementary Figure S1H). Continuous measurements were performed automatically at 6 μm intervals along the segmented border lines. OD was defined as the distance between the 2 outer wall borders, and ID was defined as the distance between the 2 inner wall borders. WLR was calculated as WT/ID78. In order to minimize the influence of cardiac pulsation on vascular measurements, an averaged image was generated using stabilized frames with a specific range of relative cardiac cycles (Fig. 5). This enabled division of the AOSLO images into 5 segments according to cardiac pulsation, and mean vascular measurements were obtained for every 5 segments. Finally, vascular measurements of 5 segments were averaged and used for analyses.


Effects of age and blood pressure on the retinal arterial wall, analyzed using adaptive optics scanning laser ophthalmoscopy.

Arichika S, Uji A, Ooto S, Muraoka Y, Yoshimura N - Sci Rep (2015)

Calculation of the mean vascular measurements correcting for cardiac pulsation.In order to minimize the influence of cardiac pulsation on vascular measurements, the cardiac cycle was synchronized to the adaptive optics (AO) videos using pulsation data obtained through a pulse oximeter attached to the subjects’ earlobes. (a) The cardiac cycle was divided into 5 segments (0–0.2, 0.2–0.4, 0.4–0.6, 0.6–0.8, and 0.8–1.0), and each video frame was assigned to the corresponding segment. (b) The images on the same relative cardiac cycle were extracted from the AO videos. (c) The images of each corresponding segment were averaged, and the vascular caliber measurements were obtained for every 5 segments. Finally, the vascular measurements of 5 segments were averaged and used for analyses.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4507481&req=5

f5: Calculation of the mean vascular measurements correcting for cardiac pulsation.In order to minimize the influence of cardiac pulsation on vascular measurements, the cardiac cycle was synchronized to the adaptive optics (AO) videos using pulsation data obtained through a pulse oximeter attached to the subjects’ earlobes. (a) The cardiac cycle was divided into 5 segments (0–0.2, 0.2–0.4, 0.4–0.6, 0.6–0.8, and 0.8–1.0), and each video frame was assigned to the corresponding segment. (b) The images on the same relative cardiac cycle were extracted from the AO videos. (c) The images of each corresponding segment were averaged, and the vascular caliber measurements were obtained for every 5 segments. Finally, the vascular measurements of 5 segments were averaged and used for analyses.
Mentions: Vascular caliber measurement was performed in zone B28, 0.5–1.0 disc diameters away from the optic disc margin using custom software known as ARIA, which was developed by Canon Inc., Tokyo, Japan. ARIA semi-automatically segments the retinal arterial wall borders in the AOSLO video. Our segmentation method consisted of preprocessing, rough central axis setting, precise central axis and vascular wall border detection, and measurement of the thickness of the vascular wall. Preprocessing consisted of the following steps. Pixel intensity values in the AOSLO video were transformed logarithmically to increase contrast in a vascular wall and decrease it in a nerve fiber layer. The AOSLO video was then stabilized using the inverse scan-line warping method. In order to obtain a high-contrast vascular wall image, frames of the stabilized AOSLO video corresponding to the specific phase range of the pulse wave were selected, and these were then smoothed using a 3-D median filter. We used a filter size of 3 × 3 × (selected frame numbers). In the next step, we manually set seed points at the center of the retinal artery along its running direction (Supplementary Figure S1A). At each seed point, the line segment that was perpendicular to the running direction of the seed points was set, and its pixel intensity profile was calculated. In the precise central axis and vascular wall border detection step, we applied a sliding linear regression filter (SLRF)3334 to the pixel intensity profile along the line segment (Supplementary Figure S1B). The SLRF method is based on the fitting of a line by linear regression that relates the pixel intensity value to the distance along the profile within a sliding window (window size: W). The precise position of the vascular central axis was identified around the seed point as a zero-cross point in the pixel intensity profile filtered by SLRF. To determine the vascular edge position, we specified the minimum point in the left side and the maximum point in the right side from the zero-cross point corresponding to the central axis. In order to determine the vascular wall border position more robustly and precisely, we applied SLRF twice to the pixel intensity profile along the line segment that was perpendicular to the central axis. Specifically, SLRF with a larger window size (W = 10 pixels) was applied first, followed by SLRF with a smaller window size (W = 4 pixels). The initial, larger window size (W = 10 pixels) was applied to establish the approximate position of the vascular wall region. In the pixel intensity profile filtered by SLRF, we could establish the approximate position of the vascular wall region by detecting the minimum point in the left side and the maximum point in the right side from the zero-cross point corresponding to the central axis (the minimum and maximum points are shown in Supplementary Figure S2C). Subsequently, we deleted outliers in the first candidate points if their Euclid distance from the central axis was less than the lower threshold (Tl)1 percentile or more than the higher threshold (Th)1percentile (Supplementary Figure S1C). In this study, Tl1 and Th1 were set as 10 and 90, respectively. The remaining first candidate points were interpolated using the natural spline interpolation method (Supplementary Figure S1D). The interpolated candidate points indicated the approximate position of the vascular wall region and were used to calculate the position of the vascular wall border by applying it to the second step (W = 4 pixels) (Supplementary Figure S1E). In the pixel intensity profile filtered by SLRF, we select the 2 nearest extremal points to the candidate point as vascular wall border candidate points (inner and outer borders are shown in Supplementary Figure S2D). Subsequently, we deleted outliers in the vascular wall border candidate points if their pixel intensity values were less than the Tl2 percentile or more than the Th2 percentile, and then deleted the remaining vascular wall border candidate points if their pixel intensity values were less than the Tl3 percentile or more than the Th3 percentile (Supplementary Figure S1F). In this study, Tl2 and Tl3 were set as 10, and Th2 and Th3 were set as 90. The remaining vascular wall border candidate points were sampled at intervals, and then interpolated using the natural spline interpolation method (Supplementary Figure S1G). Finally, we used the interpolated candidate points to define the vascular wall border in calculating the retinal arterial wall thickness (Supplementary Figure S1H). Continuous measurements were performed automatically at 6 μm intervals along the segmented border lines. OD was defined as the distance between the 2 outer wall borders, and ID was defined as the distance between the 2 inner wall borders. WLR was calculated as WT/ID78. In order to minimize the influence of cardiac pulsation on vascular measurements, an averaged image was generated using stabilized frames with a specific range of relative cardiac cycles (Fig. 5). This enabled division of the AOSLO images into 5 segments according to cardiac pulsation, and mean vascular measurements were obtained for every 5 segments. Finally, vascular measurements of 5 segments were averaged and used for analyses.

Bottom Line: WLR showed a strong correlation with age (r = 0.68, P < 0.0001), while outer diameter and inner diameter did not show significant correlation with age in the normal group (r = 0.13, P = 0.36 and r = -0.12, P =  .41, respectively).In conclusion, AOSLO provided noninvasive and reproducible arterial measurements.WLR is an early marker of morphological changes in the retinal arteries due to age and blood pressure.

View Article: PubMed Central - PubMed

Affiliation: The Department of Ophthalmology and Visual Sciences, Kyoto University Graduate School of Medicine, Kyoto 606-8507, Japan.

ABSTRACT
The wall-to-lumen ratio (WLR) of the vasculature is a promising early marker of retinal microvascular changes. Recently, adaptive optics scanning laser ophthalmoscopy (AOSLO) enabled direct and noninvasive visualization of the arterial wall. Using AOSLO, we analyzed the correlation between age and WLR in 51 normal subjects. In addition, correlations between blood pressure and WLR were analyzed in 73 subjects (51 normal subjects and 22 hypertensive patients). WLR showed a strong correlation with age (r = 0.68, P < 0.0001), while outer diameter and inner diameter did not show significant correlation with age in the normal group (r = 0.13, P = 0.36 and r = -0.12, P =  .41, respectively). In the normal and hypertensive groups, WLR showed a strong correlation with systolic and diastolic blood pressure (r = 0.60, P < 0.0001 and r = 0.65, P < 0.0001, respectively). In conclusion, AOSLO provided noninvasive and reproducible arterial measurements. WLR is an early marker of morphological changes in the retinal arteries due to age and blood pressure.

No MeSH data available.