Retinal Neurovascular and Mitochondrial Function Signatures in Response to a 24-Week Yoga Intervention in Healthy Adults

Abstract

Purpose

To evaluate changes in retinal neurovascular and mitochondrial function after a 24-week yoga intervention in healthy older adults.

Methods

Thirty participants (mean age 72 ± 6 years; 25 females, 5 males) were randomized to either a cue-based yoga program (n = 15) or traditional Hatha yoga (n = 15), with 60-minute sessions three times per week for 24 weeks. Retinal assessments were conducted at baseline and follow-up. Optical coherence tomography angiography (OCTA) was used to evaluate retinal structure and vessel density. Retinal blood flow was measured with the Retinal Function Imager, and mitochondrial function was assessed via macular flavoprotein fluorescence using the OcuMet Beacon.

Results

Significant increases were observed in retinal blood flow (2.21 to 2.72 nL/s, P = 0.002), capillary function (0.12 to 0.14 nL/s/mm, P = 0.004), and tissue perfusion (1.91 to 2.38 nL/s/mm³, P = 0.004). Macular flavoprotein fluorescence decreased (31.7 to 30.1 gsu, P = 0.041), suggesting improved mitochondrial function. No significant changes were found in vessel density, vessel length density, foveal avascular zone area, choriocapillaris density, or total retinal thickness. Between-group differences were mostly nonsignificant, except for baseline and change values in choriocapillaris density and change values in retinal thickness (P < 0.05).

Conclusions

A 24-week yoga intervention was associated with improved retinal blood flow, capillary efficiency, and tissue perfusion in healthy older adults, with indications of enhanced mitochondrial function. These findings suggest yoga may support retinal vascular and metabolic health during aging.

Yoga is a holistic spiritual practice, with improved health and well-being among its many benefits. Commonly used components for promoting health include asanas (physical postures), pranayama (controlled breathing), and meditation.¹ Incorporating regular yoga practice has been shown to be effective in improving the quality of life in different health domains such as psychiatry, oncology, and even the cardiovascular system.^2,3 These findings show the effectiveness of yoga as a promising nonpharmacological strategy to promote healthy aging and enhance quality of life.

Usage of biomarkers and systemic imaging has been vital in showing the efficacy of nonpharmacological modalities.^4–6 Although biomarkers such as blood pressure, inflammatory cytokines, and heart rate variability provide a broad overview of the body's responses to exercise or lifestyle modifications, other markers can be used to detect neurovascular and metabolic changes.⁷

The eye provides a unique and noninvasive window into neurovascular and mitochondrial health. The retina, as an extension of the brain, shares developmental and structural similarities with central nervous tissue and has a high metabolic demand supported by a dense microvascular network. As such, it is well suited to reflect systemic vascular and neuronal function.^8–10 With the development of advanced imaging techniques, including optical coherence tomography angiography (OCTA), the Retinal Function Imager, and mitochondrial function imaging, it is now possible to assess retinal structure, blood flow, and mitochondrial activity changes with high resolution and without the need for invasive procedures.^11–13

By using multimodal retinal imaging, whether the retina could serve as a sensitive biomarker for physiological adaptations associated with yoga practice can be explored. The goal of the present study was to determine whether a 24-week yoga intervention could lead to measurable changes in retinal neurovascular and mitochondrial function in healthy older adults.

Methods

Participants

This study (UM IRB 20230735) was approved by the Institutional Review Board at the University of Miami and conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent after receiving a full explanation of study procedures. Thirty healthy participants were enrolled from a research registry maintained by the Department of Kinesiology and Sports Sciences at the University of Miami, which offers wellness programs for older adults. None were actively involved in physical therapy at the time of enrollment. All participants self-reported no history of neurologic or psychiatric conditions.

Eligibility criteria included being 55 years of age or older, having no memory complaints, and scoring ≥24 on the Montreal Cognitive Assessment. Exclusion criteria included uncontrolled cardiovascular or neuromuscular disorders, cerebrovascular disease, HIV or other immune deficiencies, autoimmune or inflammatory conditions (e.g., rheumatoid arthritis, lupus), or any other serious medical illness. All participants underwent ophthalmic screening prior to enrollment, which included measurement of best-corrected visual acuity, intraocular pressure, and slit-lamp biomicroscopy. Individuals with a history of glaucoma, diabetic retinopathy, retinal detachment, or other significant ocular diseases were excluded to ensure ocular health at baseline.

After baseline testing, participants were randomly assigned to either the YogaCue or Hatha yoga intervention group. All physical and cognitive assessments were conducted at the Laboratory of Neuromuscular Research and Active Aging. Testing protocols were standardized across individuals and time points. Assessments were performed during a baseline session prior to the 24-week intervention and repeated within 2 to 3 weeks after its completion.

Sample Size and Power Considerations

To evaluate statistical sensitivity, post hoc paired power analyses were conducted in G*Power (v3.1.9.7; Heinrich-Heine-Universitat Dusseldorf, Dusseldorf, Germany) using the observed standardized changes (Cohen's dz, calculated as the linear mixed-model (LMM) coefficient [β] divided by the baseline SD). For retinal blood flow (RBF; dz = 1.09), achieved power exceeded 99% (1 − β = 0.9999; noncentrality parameter δ = 5.83). For retinal tissue perfusion (RTP; dz = 0.83), achieved power was 98.9% (δ = 4.44), and for retinal capillary function (RCF; dz = 0.96), achieved power was 99.9% (δ = 5.14). All analyses assumed α = 0.05 and a total sample size of 30 participants.

Recognizing that post hoc power is mathematically linked to observed effect sizes and P values, we emphasize the standardized coefficients and 95% confidence intervals as more informative indicators of precision. Nonetheless, the high achieved power values confirm that the present study was sufficiently sensitive to detect the observed within-subject improvements in retinal microvascular function. To guide future research, we also estimated the sample sizes required to achieve 80% power for each outcome: RBF = 10, RTP = 15, and RCF = 12. These calculations suggest that the current sample (n = 30) provided robust power for vascular outcomes.

Retinal Blood Flow Measures

Retinal blood flow was assessed using the Retinal Function Imager (RFI; Optical Imaging Ltd., Rehovot, Israel). RFI utilizes a combination of imaging and computational techniques to measure the dynamics of blood flow within the retinal vasculature and the changes and responses to various interventions.¹¹ RFI tracks erythrocyte movement within retinal vessels by analyzing the reflectance changes in the green channel of the images, providing optimal contrast for blood vessels. A series of images is captured and synced with the cardiac cycle. Blood flow velocity is calculated by measuring the displacement of erythrocyte clusters between frames.¹⁴ Participants were seated in a semi-dark room and rested for 10 minutes before imaging. Pupils were dilated with 1% tropicamide, and a single eye, preferably the right, was imaged with a 35° field of view. The same eye was used across all imaging modalities to ensure consistency. Vessel markings and flow calculations were performed using proprietary software (Browse, version 2.2.0.236), as seen in Figure 1.

Figure 1.

Retinal blood flow. (A) RBF was evaluated using the RFI within a 2.5-mm-diameter circular area centered on the fovea. Arterioles (red) and venules (pink) are displayed with their corresponding flow velocities (mm/s) overlaid. Vessel diameters were measured at the points where each vessel intersected the 2.5-mm circle. These measurements were used to calculate flow for all arterioles (marked by green dots) and venules (blue dots). Since total arteriolar inflow and venular outflow to the fovea are approximately equal, total RBF was determined by averaging the summed flow volumes of all arterioles and venules crossing the circular boundary.

To minimize variability, all imaging was conducted by a single experienced operator (J.W.) using a standardized acquisition protocol. Similarly, all image processing and vessel selection were performed by the same trained investigator following consistent criteria across participants and time points. The reliability and reproducibility of retinal blood flow velocity measurements using the RFI have been well documented. Chhablani et al.¹⁵ reported excellent intersession reproducibility of RFI-derived blood flow velocity measurements for individual retinal vessel segments, with a coefficient of variation (CV) of 10.9% and a concordance correlation coefficient (CCC) of 0.97 across two sessions 15 minutes apart. Deng et al.¹⁶ similarly demonstrated good intervisit repeatability of RFI blood flow velocities, reporting CVs around 11% and high CCCs (0.72 for arterioles and 0.67 for venules). These findings support the robustness of RFI blood flow velocity measurements, despite the use of manual vessel segment selection in our protocol.

Retinal Tissue Volume Assessment

A spectral-domain OCTA system (OptoVue, Fremont, CA, USA), a noninvasive imaging technique that provides high-resolution visualization of the retinal structure and microvasculature, was used to assess retinal tissue volume and perform layer-specific measurements. The 6 × 6-mm scans were processed using Orion image analysis software (Voxeleron LLC, Pleasanton, CA, USA), which segmented six intraretinal layers and generated corresponding volumetric data. These layers included the retinal nerve fiber layer, ganglion cell–inner plexiform layer, inner nuclear layer, outer plexiform layer, outer nuclear layer, and photoreceptor layer.¹³ Volumetric segmentation and measurement are illustrated in Figure 2.

Figure 2.

Intraretinal layers segmentation and total retinal thickness. (A) RTV was assessed using OCTA within an annular region centered on the fovea, defined by an outer diameter of 2.5 mm and an inner diameter of 0.6 mm. RTV encompassed the inner retinal layers, including the retinal nerve fiber layer (RNFL), ganglion cell–inner plexiform layer (GCIPL), inner nuclear layer (INL), and outer plexiform layer (OPL). (B) The same inner retinal layers were quantified using the AngioVue OCTA system, applying the same annular region as depicted in (A), corresponding to the circular area between the white inner boundary and the outer edge of the image. PR, photoreceptor layer; RPE, retinal pigmented epithelium.

Calculation of RTP and RCF

RTP was calculated as the total blood flow entering a 2.5-mm-diameter circular region centered on the fovea, divided by the inner retinal tissue volume, representing the average flow per mm³ of macular tissue. RCF was obtained by dividing the RBF by retinal vessel length density (RVLD), reflecting the average flow through each vessel segment.¹⁷

Retinal Vascular Measures

OCTA (AngioVue; OptoVue) was also used to assess retinal vascular network (RVN) density. The system operates at a scanning speed of 70,000 A-scans per second with an axial resolution of 5 µm. The 3 × 3-mm and 6 × 6-mm scans, centered on the fovea, were used for all participants. Image quality was evaluated and scored on a 10-point scale, and only images with a score of 7 or higher were included in the analysis. As previously described in the literature, angiographic images were analyzed using fractal dimension (Dbox; Fig. 3) via fractal analysis to quantify vessel spacing and complexity, as well as vessel density (VD).^17,18

Figure 3.

Segmentation of the OCTA image. (A) En face projection of the retinal vasculature (original OCTA image). (B) Extraction of large vessels (diameter ≥25 µm). (C) Skeletonization of small vessels obtained by subtracting large vessels from the binarized vasculature map. A circular area with a diameter of 2.5 mm centered on the fovea was analyzed to determine the fractal dimension of the retinal vasculature.

VD and vessel length density (VLD) were also quantified using the same OCTA system (AngioVue; OptoVue). OCTA images were included for analysis only if they achieved a signal strength index (SSI) of ≥7 out of 10.¹¹ En face images were then resampled to a resolution of 1024 × 1024 pixels for vessel segmentation using a custom software program developed in MATLAB (The MathWorks, Natick, MA, USA). Imaging processing specifications were described previously.¹⁹ Grayscale inversion and correction for uneven illumination were initially performed. A morphologic opening technique was then used to eliminate background noise and nonvessel elements, producing a binary image of the vessels. Vessels with diameters of 25 µm or greater were classified as large and extracted from the OCTA scans, while the rest were categorized as small vessels. The image of small vessels was subsequently skeletonized to create a map of vessel centerlines. The foveal center was automatically detected, and a ring-shaped region (annulus) with an outer diameter of 2.5 mm and an inner diameter of 0.6 mm was defined. Fractal dimension (Dbox; Fig. 3) was analyzed in the annulus of the RVN to quantify vessel spacing and complexity.¹⁸

The same annulus of the RVN was analyzed in both the vessel and skeletonized images using ImageJ (version 1.54g; National Institutes of Health, Bethesda, MD, USA) to VD, the proportion of the annular area covered by vessels, and VLD, the total vessel length in millimeters per square millimeter (mm⁻¹). Representative images and segmentation outputs for VD and VLD measurements are shown in Figure 4.

Figure 4.

Segmentation and partitioning of the retinal vascular network. (A) En face projection of the retinal vasculature with the FAZ identified. (B) Binary vessel map obtained following grayscale inversion, correction for nonuniform illumination, suppression of background noise, and morphologic filtering. (C) Definition of an annular region of interest (ROI) centered on the fovea (outer diameter: 2.5 mm; inner diameter: 0.6 mm) (D). Large vessels (diameter ≥25 µm) were isolated and extracted (E). Skeletonization of small vessels was achieved by subtracting large vessels from the binarized vasculature map. (F) Final annular ROI retaining only small vessels (diameter <25 µm) for quantitative microvascular analysis.

Choriocapillaris Density Assessment

Choriocapillaris flow density was evaluated using OCTA scans (AngioVue; Optovue) acquired over a 3 × 3-mm macula-centered area at baseline and postintervention.²⁰ The choriocapillaris slab was automatically segmented from 9 µm above to 31 µm below Bruch's membrane. Flow density, expressed as the percentage of perfused area, was quantified within a 2.5-mm-diameter circular region centered on the fovea using the device's built-in analysis software. Scans with an SSI below 7/10 or those affected by motion or segmentation artifacts were excluded (Fig. 5).

Figure 5.

Representative choriocapillaris imaging. (A) En face OCTA image of the choriocapillaris obtained from a 3 × 3-mm scan centered on the macula, illustrating the microvascular network. (B) The region of interest for choriocapillaris density quantification (highlighted in yellow) is a circular area with a 2.5-mm diameter centered on the fovea. The choriocapillaris slab was segmented from 9 µm above to 31 µm below Bruch's membrane and analyzed using optical coherence tomography angiography software.

Foveal Avascular Zone Assessment

Foveal avascular zone (FAZ) characteristics were also derived from the OCTA scans described above. The AngioVue software automatically delineated the FAZ in en face RVN slabs. Manual corrections were performed by trained graders when necessary. Metrics included FAZ area (mm²) (Fig. 4A).²⁰

Retinal Mitochondrial Function

Retinal mitochondrial function was assessed using flavoprotein fluorescence (FPF) measured by the OcuMet Beacon (OcuSciences, Ann Arbor, MI, USA). FPF captures the green autofluorescence emitted by oxidized mitochondrial flavoproteins in the retina in response to blue light excitation.¹³ Macular and optic nerve head FPF measurements were adjusted for lens signal based on age and intraocular lens compensation. The device utilizes infrared reflectance to capture images with a 60° × 21.5° field of view and FPF metabolic autofluorescence images with a 17° × 21.5° field of view. Illumination is done with an IR LED (825–870 nm) and a blue LED (458 ± 2 nm), while emitted autofluorescence is detected in the 520- to 540-nm range.²¹ Optical filters are applied to enhance the signal-to-noise ratio and reduce interference from other fluorophores in the retina and from outside the retinal plane.²² Prior to imaging, each patient had one eye pharmacologically dilated using 1.0% tropicamide. Two image locations were used: the foveal pit and the optic nerve head (ONH; Fig. 6). Image quality was graded with the OcuMet RMA software (ver. 3.0.2). Images were excluded if they were of poor quality, were off-center, or had an inadequate pupillary diameter (<3.5 mm).

Figure 6.

Flavoprotein fluorescence (FPF) analysis of the macula and optic nerve head. Representative postprocessed output from the OcuMet Beacon system (RMA software version 3.0.2) displays macular and ONH FPF analysis along with corresponding image quality metrics. (A) The analyzed macular region and ONH are delineated by a green circle and square, respectively. Infrared image quality (IR IQ) is rated on a scale from 0 to 100, with scores of 0 to 30 classified as poor, 30 to 70 as acceptable, and 70 to 100 as good. (B) Macular FPF image quality (Mac FPF IQ) is similarly evaluated using the same scale. Pupil diameter (in millimeters) is indicated, with only eyes measuring greater than 3.5 mm included in the final analysis—a whole-image FPF signal heatmap shows a 4.8-mm macular analysis zone outlined in green. The mean retinal FPF value (“Retina”) is calculated over an approximately 5.1-mm diameter area centered on the fovea. Subject age is estimated based on lens autofluorescence. (C) A corresponding localized hotspot heatmap highlights regions of increased metabolic activity within the macula (D). The ONH is marked by a green square region of interest, with autofluorescence photobleaching factor measurements taken from the annulus area between two red circles outlining the optic nerve rim. Optic disc FPF image quality (Dsc FPF IQ) is scored from 0 to 100: a score of 0 to 30 denotes poor quality, 31 to 70 is acceptable, and above 70 indicates good quality. (E) A line graph shows the FPF profile as measured from temporal to nasal to temporal around the optic nerve rim in both eyes, highlighting spatial differences in mitochondrial dysfunction within the ONH.

Yoga Training Program

Participants attended yoga classes three times per week for 24 weeks, with each session lasting 60 minutes, led by certified instructors. To meet study requirements, participants were required to attend at least 85% of all sessions, and the average attendance rate across both groups was 91% per participant. The YogaCue group followed a novel, cognitively engaging program developed by two of the authors (JFS and KJM), which incorporated multidirectional movement, progressive speed changes, and auditory-visual cues designed to target working memory, cognitive flexibility, and executive function.²³ The approach emphasized rapid transitions, psychomotor training, and increasing cognitive load through monthly progressions to enhance physical and mental performance. Participants were also encouraged to view sessions as game-like to boost engagement.

In contrast, the Hatha group practiced traditional Hatha yoga, focusing on slower transitions, postural alignment, and breathwork (Fig. 7). Sessions emphasized static poses, including foundational movements and progressively challenging variations to maintain physical overload throughout the program. Hatha yoga participants performed similar exercises to YogaCue but in a less structured order, at a slower pace, with fewer repetitions, and at a lower intensity.

Figure 7.

Yoga postures are included in the training. Selected yoga postures incorporated into the training include Warrior I, Warrior II, Mountain, Triangle, Tree, and Airplane poses, chosen for their emphasis on enhancing balance, strength, and stability.

Relative intensity in both groups was designed to increase monthly. This was achieved by shortening posture hold and transition durations, performing sequences at a faster pace, and increasing the number of repetitions per session. For example, posture holds and transitions were initially measured in breaths (e.g., five breaths per pose, three transition breaths), and these counts decreased over time, making movements faster and more challenging. Sequence repetitions also increased progressively; during the first month, each flow sequence was performed one to two times per session, with additional repetitions added each month to increase workload.

The YogaCue program included four structured flow sequences: flow 1 (Warrior I, Warrior II, Reverse Warrior, Warrior II, Crescent Lunge, Airplane Lunge, Chair), flow 2 (Warrior I, Warrior II, Reverse Warrior, Warrior II, Side Angle, Warrior II, Triangle, Warrior II, Crescent Lunge, Airplane Lunge, Chair), flow 3 (Warrior I, Warrior II, Reverse Warrior, Warrior II, Five-Pointed Star, Prayer Squat repeated three times, Five-Pointed Star, Wide-Legged Forward Fold, Warrior II, Crescent Lunge, Airplane Lunge, Chair), and flow 4 (Warrior I, Warrior II, Reverse Warrior, Crescent Lunge, Airplane Lunge, Crescent Lunge, Tree, Tall Mountain).

No direct measurements of exercise intensity, such as metabolic equivalents or heart rate, were collected. However, the progressive adjustments in posture duration, transitions, and repetitions were intended to systematically increase relative intensity over the 24-week intervention. While the metabolic cost of each session was not assessed in this study, the high-speed, interval-based structure of the YogaCue program was designed to reflect the intensity pattern of high-speed interval yoga previously reported by Potiaumpai et al.,²⁴ where faster transitions and shorter posture durations produced significantly greater energy expenditure than traditional Hatha yoga (mean difference ± SE = 18.6 ± 1.9 kcal; P < 0.01). This framework guided the progression and differentiation of the YogaCue intervention compared with traditional yoga practice.

Statistical Analysis and Imaging Analysis

All statistical analyses were performed using SPSS version 27.0 (SPSS, Chicago, IL, USA) and STATA version 18.5 (StataCorp, College Station, TX, USA). Normality of distributions was assessed with the Shapiro–Wilk test. Continuous variables are presented as mean ± SD.

Longitudinal changes in retinal vascular, structural, and mitochondrial parameters were evaluated using linear mixed-effects models in STATA, controlling for age, sex, and lens status. Between-group comparisons of the two yoga subgroups (YogaCue versus Hatha yoga) were conducted in SPSS: the independent-samples t-test was used for normally distributed variables, while the Mann–Whitney U test was applied to nonnormally distributed variables (RBF, RTP, RCF, and macular FPF). Categorical variables (e.g., sex) were compared with the χ² test.

Correlations between vascular, structural, and metabolic parameters were assessed using the Pearson correlation for normally distributed data and Spearman rank correlation for nonnormally distributed data. Statistical significance was set at P < 0.05 for all analyses.

Results

Thirty participants were included in this study, with a mean age of 72 ± 6 years (range: 56–84 years), comprising 25 females and 5 males (Table 1). Participants were randomly stratified into YogaCue (n = 15; 12 females/3 males; mean age = 73 ± 7 years) and Hatha yoga (n = 15; 13 females/2 males; mean age = 72  ±  5 years) groups. There were no significant differences between the groups in age (t(28) = 0.72, P = 0.48) or sex (χ², P = 0.62).

Table 1.

Demographic and Clinical Characteristics of Participants at Baseline (V1) and Follow-Up (V2)

Variable	V1 (Baseline), Mean ± SD	V2 (Follow-Up), Mean ± SD	P Value
Age (years)	72 ± 6	73 ± 6	< 0.001
Gender, female/male (n)		25/5
HR	65 ± 9	67 ± 9	0.23
Systolic BP (mm Hg)	126.4 ± 14.2	123.0 ± 10.9	0.19
Diastolic BP (mm Hg)	76.9 ± 7.9	75.3 ± 7.0	0.22
BMI (kg/m2)	28.0 ± 5.0	27.6 ± 4.7	0.14
Weight (kg)	75.2 ± 15.6	74.2 ± 15.0	0.14
IOP (mm Hg)	13.7 ± 1.6	14.1 ± 2.3	0.51

BMI, body mass index; BP, blood pressure; HR, heart rate; IOP, intraocular pressure.

Bold font indicates statistical significance.

Following the 24-week yoga intervention, significant increases were observed in RBF, RTP, and RCF, while macular FPF significantly decreased (Table 2; Fig. 8). No significant changes were detected in retinal vessel density (RVD), RVLD, fractal dimension of the retinal vascular network (FD), foveal avascular zone (FAZ) area, choriocapillaris density (CCD), ONH FPF, or total retinal thickness (TRT).

Table 2.

Retinal Vascular, Structural, and Mitochondrial Parameters at Baseline and 24-Week Follow-Up

Variable	N	Baseline, Mean ± SD	Follow-Up, Mean ± SD	LMM β	95% CI	SE	z	P
RBF (nL/s)	30	2.21 ± 0.47	2.72 ± 0.93	0.51	0.19 to 0.83	0.16	3.14	0.002
RTP (nL/s/mm³)	28	1.91 ± 0.54	2.38 ± 0.90	0.446	0.15 to 0.75	0.153	2.92	0.004
RCF (nL/s/mm)	28	0.115 ± 0.026	0.141 ± 0.050	0.025	0.008 to 0.43	0.009	2.85	0.004
Macular FPF (gsu)	24	31.7 ± 10.2	30.1 ± 9.8	− 1.60	−3.14 to –0.06	0.79	−2.04	0.04
ONH FPF (gsu)	22	37.79 ± 11.2	37.14 ± 10.9	−0.64	−2.25 to 0.96	0.82	−0.79	0.43
RVD (%)	28	32.41 ± 1.2	32.20 ± 1.1	−0.21	−0.59 to 0.17	0.19	−1.09	0.28
RVLD (mm−1)	28	19.39 ± 1.0	19.32 ± 0.9	−0.071	−0.42 to 0.28	0.178	−0.4	0.69
FAZ area (mm2)	29	0.246 ± 0.09	0.243 ± 0.10	−0.003	−0.01 to 0.01	0.005	−0.67	0.50
CCD (%)	29	63.99 ± 4.5	64.31 ± 3.3	0.003	−0.01 to 0.02	0.007	0.46	0.65
FD (Dbox)	28	1.796 ± 0.009	1.797 ± 0.009	0.001	−0.003 to 0.005	0.002	0.56	0.58
TRT 3 mm (mm)	28	2.131 ± 0.14	2.124 ± 0.13	−0.008	−0.02 to 0.005	0.006	−1.22	0.22
TRT 6 mm (mm)	28	7.845 ± 0.36	7.815 ± 0.36	−0.030	−0.07 to 0.01	0.020	−1.47	0.14

Linear mixed models (LMMs) results indicate change over time. Bold font indicates statistical significance.

Figure 8.

Retinal vascular, structural, and mitochondrial function responses to 24-week Yoga exercise. (A–C) Significant increases in RBF, RTP, and RCF were observed from baseline to follow-up (P < 0.05). (D–H) Changes in vascular and structural measurements did not reach statistical significance. (I) A significant reduction in macular FPF was noted (P = 0.04). (J) No significant changes were found in TRT within the 2.5-mm and 6-mm circular areas. Bold font indicates statistical significance.

Given the robust changes observed in RBF, we first describe its detailed trajectory over the 24-week intervention. At baseline, the mean RBF was 2.21 ± 0.47 nL/s, rising to 2.72 ± 0.93 nL/s at follow-up. Linear mixed-effects analysis confirmed a significant increase over time (β = 0.51; 95% confidence interval [CI], 0.19–0.83, SE = 0.16, z = 3.14, P = 0.002), accounting for age, sex, and lens status. Age was positively associated with RBF (β = 0.03, P = 0.04), while males had lower RBF than females (β = −0.6, P = 0.004); lens status was not significant (β = −0.007, P = 0.97). Predictive margins, representing the estimated mean RBF at each visit, were 2.23 nL/s (95% CI, 2.08–2.38) at baseline and 2.74 nL/s (95% CI, 2.41–3.07) at follow-up.

Beyond RBF, RTP and RCF also showed significant increases over the 24-week intervention. RTP increased from 1.91 ± 0.54 nL/s/mm³ at baseline to 2.38 ± 0.90 nL/s/mm³ at follow-up (β = 0.45; 95% CI, 0.15–0.75; P = 0.004), while RCF rose from 0.115 ± 0.026 nL/s/mm to 0.141 ± 0.050 nL/s/mm (β = 0.025; 95% CI, 0.008–0.043; P = 0.004), as confirmed by linear mixed-effects models adjusting for age, sex, and lens status. Macular FPF decreased from 31.7 ± 10.2 gsu to 30.1 ± 9.8 gsu (β = −1.60; 95% CI, −3.13 to −0.07; P = 0.041), suggesting a reduction in mitochondrial oxidative stress.

To provide a measure of effect size, we calculated standardized changes by dividing the LMM coefficient (β) by the baseline standard deviation of each parameter, analogous to Cohen's d. This allows the magnitude of change to be interpreted relative to baseline variability. The standardized changes were as follows: RBF increased by 1.09 SD, RTP increased by 0.83 SD, RCF increased by 0.96 SD, and macular FPF decreased by 0.16 SD. These results indicate that the yoga intervention produced substantial improvements in retinal microvascular function, with more modest but detectable changes in mitochondrial metabolic function.

No significant changes were observed in RVD, RVLD, FD, FAZ area, CCD, optic nerve head FPF, or TRT (all P > 0.05), indicating that the 24-week yoga intervention predominantly affected microvascular perfusion and capillary function rather than retinal structure.

Between-group comparisons of the two yoga modalities revealed no significant differences for most parameters. However, baseline CCD and the changes in CCD and TRT showed significant differences between the two groups (all P < 0.05; Table 3).

Table 3.

Retinal Vascular, Structural, and Mitochondrial Function Parameters at Both Baseline and Follow-Up for Each Subgroup

Variable	Sample Size	Group	Baseline, Mean ± SD	P Value	Sample Size	Follow-Up, Mean ± SD	P Value	Delta, Mean ± SD
RBF (nL/s)	15	YogaCue	2.22 ± 0.47	0.68	15	2.59 ± 0.74	0.54	0.37 ± 0.88
	15	Hatha yoga	2.21 ± 0.54		15	2.85 ± 1.09		0.63 ± 1.10
RTP (nL/s/mm3)	14	YogaCue	1.90 ± 0.54	0.45	14	2.29 ± 0.68	0.98	0.39 ± 0.82
	14	Hatha yoga	1.92 ± 0.67		15	2.48 ± 1.12		0.54 ± 1.05
RCF (nL/s/mm)	14	YogaCue	0.12 ± 0.03	0.40	14	0.14 ± 0.04	0.87	0.02 ± 0.05
	14	Hatha yoga	0.11 ± 0.03		14	0.15 ± 0.05		0.03 ± 0.06
Macular FPF (gsu)	13	YogaCue	32.4 ± 10.6	0.48	12	29.4 ± 10.8	0.68	−0.8 ± 6.4
	13	Hatha yoga	32.6 ± 9.6		15	30.5 ± 8.5		−0.3 ± 8.4
ONH FPF (gsu)	12	YogaCue	37.7 ± 10.7	0.81	11	37.5 ± 14.8	0.84	−0.7 ± 3.7
	12	Hatha yoga	36.7 ± 6.7		13	38.4 ± 6.2		0.3 ± 3.9
VLD (mm−1)	14	YogaCue	19.1 ± 1.0	0.13	15	19.3 ± 1.0	0.63	0.2 ± 1.0
	14	Hatha yoga	19.7 ± 1.2		15	19.4 ± 0.9		−0.3 ± 1.0
VD (%)	14	YogaCue	32.2 ± 1.2	0.40	15	32.1 ± 1.1	0.57	−0.1 ± 1.1
	14	Hatha yoga	32.6 ± 1.3		15	32.3 ± 1.0		−0.3 ± 1.1
FD (Dbox)	14	YogaCue	1.794 ± 0.009	0.42	15	1.798 ± 0.008	0.84	0.00 ± 0.01
	15	Hatha yoga	1.798 ± 0.010		14	1.797 ± 0.010		0.00 ± 0.01
CCD (%)	15	YogaCue	66.4 ± 3.8	0.003	15	65.0 ± 3.6	0.24	−1.4 ± 3.8
	14	Hatha yoga	61.7 ± 3.9		14	63.5 ± 2.8		1.8 ± 3.5
FAZ area (mm2)	15	YogaCue	0.26 ± 0.09	0.60	15	0.25 ± 0.07	0.77	−0.01 ± 0.02
	14	Hatha yoga	0.24 ± 0.11		14	0.24 ± 0.12		0.00 ± 0.02
Total retinal volume (6 mm)	14	YogaCue	7.78 ± 0.36	0.37	14	7.71 ± 0.29	0.11	−0.06 ± 0.11
	14	Hatha yoga	7.90 ± 0.40		15	7.93 ± 0.39		0.02 ± 0.05
Total retinal volume (2.5 mm)	14	YogaCue	2.13 ± 0.14	0.97	14	2.11 ± 0.12	0.54	−0.02 ± 0.03
	14	Hatha yoga	2.13 ± 0.14		15	2.14 ± 0.13		0.00 ± 0.02

Bold font indicates statistical significance.

Changes in microvascular and metabolic function were correlated: alterations in FAZ area were positively associated with changes in macular FPF (ρ = 0.49, P = 0.01), suggesting a link between microvascular remodeling and mitochondrial metabolism. This association was attenuated and nonsignificant after adjusting for age and sex (ρ = 0.32, P = 0.18).

At baseline, TRT in the 6-mm circular area correlated with RVD (r = 0.40, P = 0.03; Fig. 9A) and RVLD (r = 0.54, P = 0.003; Fig. 9B). Postintervention, TRT in the 2.5-mm region was correlated with RVD (r = 0.50, P = 0.006), RVLD (r = 0.55, P = 0.02), and FD (r = 0.43, P = 0.02). After adjusting for age and sex, TRT remained significantly associated with RVLD both at baseline (r = 0.46, P = 0.02) and at follow-up (r = 0.40, P = 0.04).

Figure 9.

Relations between TRT, RVD, and RVLD. Scatterplots showing the relationships between TRT within a 6-mm diameter circle centered on the fovea and both RVD and RVLD at baseline.

Discussion

To our knowledge, this is the first study to evaluate the effects of a 24-week yoga intervention on retinal neurovascular and mitochondrial function in healthy older adults, demonstrating measurable improvements in both domains. Retinal blood flow, as well as metrics reflecting capillary efficiency and tissue perfusion, increased after the intervention, suggesting enhanced vascular performance. Importantly, a statistically significant reduction in macular FPF was also observed, suggesting improved mitochondrial efficiency and reduced oxidative stress. In contrast, structural features of the retina, including vessel density, vascular complexity, and overall retinal volume, remained essentially unchanged. Comparisons between different yoga groups revealed minimal differences, except in a few baseline and delta measures related to choriocapillaris density and retinal thickness. Overall, the findings suggest that yoga improves physiological health, evidently through improved blood flow and metabolic function in the retina.

Following the 24-week yoga intervention, increases were observed mainly in the circulation domain, including RBF, RTP, and RCF. These changes suggest that yoga may enhance circulation efficiency and overall vascular health. RBF reflects the volume of blood delivered to the retinal tissue.²⁵ An increase in RBF suggests better delivery of oxygen and nutrients to the highly metabolic tissue, such as the retina. Yoga may enhance RBF through multiple pathways, including improved autonomic balance, reduced sympathetic tone, and enhanced endothelial function.²⁶ Regular yoga practice has been shown to promote vasodilation and reduce arterial stiffness, partly by increasing nitric oxide bioavailability and reducing systemic inflammation.^27,28 These effects may contribute to improved vascular tone and perfusion in both systemic and ocular circulations. RTP is the blood flow normalized by the volume of retinal tissue.^17,25 This measure reflects how effectively the tissue receives oxygen and metabolic substrates. An increase in RTP suggests improved blood supply to meet the metabolic support to the tissue, which is critical for maintaining cell function and health.²⁹ Enhanced tissue perfusion may also indicate better matching of blood supply to metabolic demand, a process known as neurovascular coupling.³⁰ Yoga has been associated with improved cerebral perfusion and better regulation of microcirculation, possibly through modulation of vascular growth factors and enhanced parasympathetic activity.^31–33

Furthermore, RCF, as blood flow per unit of vessel length, provides a measure of functional efficiency of the vessel bed, mainly the capillaries.¹⁷ An increase in RCF implies that existing vessels can deliver more blood, reflecting improved microvascular responsiveness to the demand. This could be driven by enhanced endothelial-dependent vasodilation, greater capillary recruitment, and improved circulation dynamics, all of which have been observed in individuals engaging in regular yoga or other moderate-intensity exercise.³⁴

Indeed, yoga incorporates slow breathing, isometric holds, and meditative components, which are known to enhance baroreceptor sensitivity, reduce cortisol levels, and modulate the hypothalamic–pituitary–adrenal axis.^26,35,36 These systemic changes can reduce oxidative stress and support mitochondrial health, thereby preserving the integrity of retinal neurons and capillary networks.³⁷ Together, the improvements in RBF, RTP, and RCF observed in the present study suggest that yoga can improve the health of the neurovascular unit, as evident in the retina, by enhancing vascular function and metabolic efficiency. These findings align with broader evidence showing that mind–body interventions can induce favorable physiological adaptations in the body, including the brain and its extension (i.e., retina).^27,38

The magnitude of these changes, as indicated by standardized effect sizes previously reported, suggests that the improvements in retinal microvascular function are substantial, while the reduction in macular FPF reflects a modest yet meaningful enhancement of mitochondrial efficiency. Collectively, these effects indicate that a 24-week yoga intervention may have physiologically and potentially clinically relevant benefits for retinal vascular and metabolic health in older adults.

In parallel with these vascular changes, a significant reduction in macular FPF was observed following the 24-week yoga intervention, suggesting a potential improvement in mitochondrial function. FPF reflects the level of oxidized flavoproteins in the electron transport chain and serves as a sensitive indicator of oxidative stress and mitochondrial inefficiency, particularly in metabolically active tissues such as the retina.^22,39 Elevated FPF has been associated with increased production of reactive oxygen species, disrupted mitochondrial membrane potential, and reduced ATP synthesis.⁴⁰ The reduction in FPF seen after yoga practice may reflect improved redox homeostasis and enhanced mitochondrial efficiency. Yoga has been shown to decrease systemic oxidative stress markers, such as malondialdehyde, and increase the activity of endogenous antioxidants like superoxide dismutase and glutathione peroxidase.^41,42 These antioxidant effects may reduce mitochondrial oxidative burden and preserve cell function.

Additionally, yoga has been associated with improved mitochondrial respiration and ATP production, potentially through the upregulation of mitochondrial biogenesis and more efficient energy metabolism.^42,43 Autonomic regulation also plays a role; by enhancing parasympathetic tone and reducing sympathetic overactivity, yoga lowers systemic metabolic demand and catecholamine-driven mitochondrial stress.^28,36,44 Furthermore, the anti-inflammatory effects of yoga may also support mitochondrial function by mitigating cytokine-induced mitochondrial dysfunction.⁴⁵ Although these cellular benefits are well documented, the American College of Sports Medicine notes that traditional forms like Hatha yoga are typically low to moderate in intensity and may not meet aerobic fitness thresholds.⁴⁶ Thus, yoga may support mitochondrial health primarily through nonaerobic mechanisms, contributing to reduced FPF.

In addition to the key findings, a notable relationship was observed between changes in FAZ area and macular FPF, suggesting a possible link between microvascular remodeling and mitochondrial function. FAZ enlargement usually signals capillary dropout or ischemia, exacerbating mitochondrial dysfunction.^47,48 Stabilization or reduction of FAZ during intervention may indicate improved vascular integrity and reduced metabolic stress. The attenuation of FAZ–FPF associations after adjusting for age and sex likely reflects established demographic effects on retinal vasculature and metabolism.^18,39,49 Aging, for example, reduces deep retinal perfusion and vessel density, thins retinal layers, and increases mitochondrial stress (FPF), particularly at the optic nerve head, where the normal range of FPF values also widens.^13,18 As a result, many structure–function relationships disappear after accounting for age.¹³ These factors contribute to variability in vascular and mitochondrial responses beyond the exercise intervention.

Moreover, baseline retinal volume correlated with vessel density and length density, suggesting that under normal conditions, capillary networks support tissue mass. However, these associations weakened after adjusting for age and sex, highlighting demographic influences on retinal structure–vascular relationships.

The observed improvements in retinal blood flow and metabolic efficiency suggest that yoga may offer therapeutic potential for populations at risk of vascular and mitochondrial dysfunction. Retinal diseases such as age-related macular degeneration, diabetic retinopathy, and glaucoma are all characterized by compromised microcirculation and oxidative stress, mechanisms that may be modifiable through regular mind–body practices. Additionally, given the retina's embryological and functional continuity with the brain, these findings raise the possibility of broader neurologic benefits, particularly in aging populations or individuals with early cognitive decline.

Future studies should extend this work to clinical cohorts to evaluate whether yoga can serve as an adjunctive intervention to preserve visual and neurologic function. Comparative studies involving aerobic exercise or other structured interventions could help identify the most effective strategies for enhancing neurovascular and mitochondrial health. Longitudinal designs with larger, more diverse samples and the inclusion of systemic biomarkers would further clarify the mechanisms of action and improve the generalizability of these findings.

This study has several limitations. A major limitation is the absence of a nonexercising control group, which limits our ability to draw definitive causal conclusions regarding the effects of yoga. Nevertheless, RBF, the primary outcome, is known to remain relatively stable over time in healthy adults. For instance, Zhang et al.⁵⁰ reported no significant changes in RBF and RTP over 8 weeks in a control cohort using the same measurement device. Likewise, our prior work with OCTA in a 24-week nonintervention control group showed no significant changes in retinal capillary density, supporting the reproducibility of these measures.⁵¹ Taken together, the significant increases in RBF, retinal tissue perfusion, and capillary function observed in this study are unlikely to reflect normal temporal variation and are more likely attributable to the yoga intervention.

Moreover, this study focused on healthy older adults, which is essential for understanding aging but also presents some challenges. Older adults often show more variation in vascular and mitochondrial function, making it harder to detect consistent effects.^13,18 Age-related eye changes, such as lens opacity or small pupil size, may also affect imaging quality. Some participants may have had physical limitations that reduced their ability to fully participate in yoga, leading to differences in how much benefit they received. Finally, while the retina offers a view of vascular and metabolic health, it only gives an indirect picture of the body as a whole. Future studies should include more participants, a parallel control group, and additional systemic measures.

Conclusions

This pilot study is the first to evaluate the effects of a structured 24-week yoga intervention on retinal neurovascular and mitochondrial function in healthy older adults using noninvasive imaging techniques. Our findings indicate that yoga may promote improvements in retinal blood flow, tissue perfusion, and capillary efficiency, along with a reduction in mitochondrial oxidative stress, as suggested by decreased flavoprotein fluorescence. These physiological changes suggest enhanced microvascular and metabolic function in the retina, an accessible extension of the central nervous system and a potential biomarker of systemic health.

While these results are encouraging, the absence of a randomized control group limits definitive conclusions about causality. Nevertheless, this study introduces retinal imaging as a novel and objective approach for detecting early physiological responses to mind–body interventions.

Further research, including randomized controlled trials with larger and more diverse populations, is needed to confirm these associations, explore systemic correlates, and assess the relevance of yoga-based interventions in populations at risk for vascular and neurodegenerative conditions. These early insights support the growing potential of the retina as a noninvasive window into the effects of mind–body practices on neurovascular and mitochondrial health during aging.