IMAGE EVALUATION OF ARTIFICIAL INTELLIGENCE-SUPPORTED OPTICAL COHERENCE TOMOGRAPHY ANGIOGRAPHY IMAGING USING OCT-A1 DEVICE IN DIABETIC RETINOPATHY

Pseudoaveraging for denoising of OCT angiography: a deep learning approach for image quality enhancement in healthy and diabetic eyes

BACKGROUND

This study aimed to develop a deep learning (DL) algorithm that enhances the quality of a single-frame enface OCTA scan to make it comparable to 4-frame averaged scan without the need for the repeated acquisitions required for averaging.

METHODS

Each of the healthy eyes and eyes from diabetic subjects that were prospectively enrolled in this cross-sectional study underwent four repeated 6 × 6 mm macular scans (PLEX Elite 9000 SS-OCT), and the repeated scans of each eye were co-registered to produce 4-frame averages. This prospective dataset of original (single-frame) enface scans and their corresponding averaged scans was divided into a training dataset and a validation dataset. In the training dataset, a DL algorithm (named pseudoaveraging) was trained using original scans as input and 4-frame averages as target. In the validation dataset, the pseudoaveraging algorithm was applied to single-frame scans to produce pseudoaveraged scans, and the single-frame and its corresponding averaged and pseudoaveraged scans were all qualitatively compared. In a separate retrospectively collected dataset of single-frame scans from eyes of diabetic subjects, the DL algorithm was applied, and the produced pseudoaveraged scan was qualitatively compared against its corresponding original.

RESULTS

This study included 39 eyes that comprised the prospective dataset (split into 5 eyes for training and 34 eyes for validating the DL algorithm), and 105 eyes that comprised the retrospective test dataset. Of the total 144 study eyes, 58% had any level of diabetic retinopathy (with and without diabetic macular edema), and the rest were from healthy eyes or eyes of diabetic subjects but without diabetic retinopathy and without macular edema. Grading results in the validation dataset showed that the pseudoaveraged enface scan ranked best in overall scan quality, background noise reduction, and visibility of microaneurysms (p < 0.05). Averaged scan ranked best for motion artifact reduction (p < 0.05). Grading results in the test dataset showed that pseudoaveraging resulted in enhanced small vessels, reduction of background noise, and motion artifact in 100%, 82%, and 98% of scans, respectively. Rates of false-positive/-negative perfusion were zero.

CONCLUSION

Pseudoaveraging is a feasible DL approach to more efficiently improve enface OCTA scan quality without introducing notable image artifacts.

Inference of Capillary Nonperfusion Progression on Widefield OCT Angiography in Diabetic Retinopathy

Purpose

The purpose of this study was to explore the spatial patterns of the nonperfusion areas (NPAs) on widefield optical coherence tomography angiography (OCTA) images in diabetic retinopathy (DR) and to investigate their associations with NPA progression and DR severity.

Methods

We prospectively enrolled 201 eyes from 158 patients with DR. Widefield images were obtained using a swept-source OCTA device (Xephilio OCT-S1), followed by the creation of 20-mm (1614 pixels) en face images. Nonperfusion squares (NPSs) were defined as 10 × 10-pixel squares without retinal vessels. Eyes with high-dimensional spatial data were mapped onto a two-dimensional space using the uniform manifold approximation and projection algorithm and divided by clustering. The patterns of NPA distribution were statistically compared between clusters.

Results

All eyes were mapped onto a two-dimensional space and divided into six clusters based on the similarity of NPA distribution. Eyes in clusters 1 and 2 had minimal and small NPAs, respectively. Eyes in clusters 3 and 4 exhibited NPAs in the temporal and inferotemporal regions, respectively. Eyes in cluster 5 displayed NPAs in both superonasal and inferonasal areas. The unique NPA distributions in each cluster encouraged us to propose eight possible pathways of NPA progression. DR severity was not equal between clusters (P < 0.001), for example, 8 (15.7%) of 51 eyes and 15 (65.2%) of 23 eyes had PDR in clusters 1 and 5, respectively.

Conclusions

Dimensionality reduction and subsequent clustering based on the NPA distribution on widefield OCTA enabled the inference of possible NPA progression in DR.

Assessing the Sensitivity of OCT-A Retinal Vasculature Metrics

Purpose

The purpose of this study was to examine the sensitivity of quantitative metrics of the retinal vasculature derived from optical coherence tomography angiography (OCT-A) images.

Methods

Full retinal vascular slab OCT-A images were obtained from 94 healthy participants. Capillary loss, at 1% increments up to 50%, was simulated by randomly removing capillary segments (1000 iterations of randomized loss for each participant at each percent loss). Thirteen quantitative metrics were calculated for each image: foveal avascular zone (FAZ) area, vessel density, vessel complexity index (VCI), vessel perimeter index (VPI), fractal dimension (FD), and parafoveal intercapillary area (PICA) measurements with and without the FAZ (mean PICA, summed PICA, PICA regularity, and PICA standard deviation [PICA SD]). The sensitivity of each metric was calculated as the percent loss at which 80% of the iterations for a participant fell outside of two standard deviations from the sample's normative mean.

Results

The most used OCT-A metrics, FAZ area and vessel density, were not significantly different from normative values until 27.69% and 16.00% capillary loss, respectively. Across the remaining metrics, metric sensitivity ranged from 6.37% (PICA SD without FAZ) to 39.78% (Summed PICA without FAZ).

Conclusions

The sensitivity of vasculature metrics for detecting random capillary loss varies substantially. Further efforts simulating different patterns of capillary loss are needed for comparison. Additionally, mapping the repeatability of metrics over time in a normal population is needed to further define metric sensitivity.

Translational Relevance

Quantitative metrics vary in their ability to detect vascular abnormalities in OCT-A images. Metric choice in screening studies will need to balance expected capillary abnormalities and the quality of the OCT-A images being used.

Artificial Intelligence-Aided Optical Imaging for Cancer Theranostics

The use of artificial intelligence (AI) to assist biomedical imaging have demonstrated its high accuracy and high efficiency in medical decision-making for individualized cancer medicine. In particular, optical imaging methods are able to visualize both the structural and functional information of tumors tissues with high contrast, low cost, and noninvasive property. However, no systematic work has been performed to inspect the recent advances on AI-aided optical imaging for cancer theranostics. In this review, we demonstrated how AI can guide optical imaging methods to improve the accuracy on tumor detection, automated analysis and prediction of its histopathological section, its monitoring during treatment, and its prognosis by using computer vision, deep learning and natural language processing. By contrast, the optical imaging techniques involved mainly consisted of various tomography and microscopy imaging methods such as optical endoscopy imaging, optical coherence tomography, photoacoustic imaging, diffuse optical tomography, optical microscopy imaging, Raman imaging, and fluorescent imaging. Meanwhile, existing problems, possible challenges and future prospects for AI-aided optical imaging protocol for cancer theranostics were also discussed. It is expected that the present work can open a new avenue for precision oncology by using AI and optical imaging tools.

Wide-field swept-source OCT angiography (23 × 20 mm) for detecting retinal neovascularization in eyes with proliferative diabetic retinopathy

PURPOSE

Xephilio OCT-S1 can capture single-acquisition 23 × 20-mm wide-field swept-source optical coherence tomography angiography (SS-OCTA) images and high-resolution images using artificial intelligence. We aimed to evaluate the ability of wide-field SS-OCTA in the detection of retinal neovascularizations (NVs) in eyes with proliferative diabetic retinopathy (PDR).

METHODS

This retrospective study included 64 eyes of 36 patients (age, 57 ± 10 years; 10 female, 26 male) with PDR. All patients underwent a comprehensive ophthalmological examination, including fluorescein angiography (FA), as well as fovea- and disc-centered 23 × 20-mm OCTA imaging (A-scan/B-scan, 928/807). We compared and examined the number of NV sites identified using conventional methods (merging the findings from biomicroscopy/color fundus photography, FA) and the number of NV sites identified using vitreoretinal interface and superficial retinal slabs of wide-field SS-OCTA images, including the position of NVs (nasal upper, nasal lower, temporal upper, temporal lower, or disc).

RESULTS

We identified 168 NVs (32/40/45/35/16, in the abovementioned order) using the conventional method. Fovea-centered 23 × 20-mm OCTA images revealed 162 (96%) NVs (27/39/45/35/16). This method tended to miss nasal NV. In contrast, disc-centered 23 × 20-mm OCTA images identified nearly all NVs, detecting 166 (99%) NVs (32/40/44/34/16) in total. All NVs could be visualized using two wide-field OCTA images: fovea- and disc-centered.

CONCLUSION

Wide-field (23 × 20 mm) SS-OCTA-especially disc-centered-using Xephilio OCT-S1 identified nearly all NVs in eyes with PDR, with a single acquisition, thereby demonstrating its potential clinical application.

Deep Learning-Based Noise Reduction Improves Optical Coherence Tomography Angiography Imaging of Radial Peripapillary Capillaries in Advanced Glaucoma

PURPOSE

We applied deep learning-based noise reduction (NR) to optical coherence tomography-angiography (OCTA) images of the radial peripapillary capillaries (RPCs) in eyes with glaucoma and investigated the usefulness of this method as an objective analysis of glaucoma.

METHODS

This cross-sectional study included 118 eyes of 94 open-angle glaucoma patients (male/female = 38/56, age: 56.1 ± 10.3 years). We used OCTA (OCT-HS100, Canon) and built-in software (RX software, v. 4.5) to perform NR and calculate RPC vessel area density (VAD) and skeleton vessel length density (VLD). We also examined NR's effect on reproducibility. Finally, we assessed the vascular structure (PRCs)/function relationship at different glaucoma stages with Spearman's correlation.

RESULTS

Regardless of NR, RPC parameters had excellent coefficients of variation (1.7-4.1%) in glaucoma patients and controls, and mean deviation (MD) was significantly correlated with VAD (NR: r = 0.835, p < 0.001; non-NR: r = 0.871, p < 0.001) and VLD (NR: r = 0.829, p < 0.001; non-NR: r = 0.837, p < 0.001). For mild, moderate, and advanced glaucoma, the correlation coefficients between MD and VLD were 0.366 (p = 0.028) 0.081 (p = 0.689), and 0.427 (p = 0.017) with NR and 0.405 (p = 0.014), 0.184 (p = 0.360), and 0.339 (p = 0.062) without NR, respectively.

CONCLUSION

Denoised RPC images might have the potential for a closer structural/functional relationship, in which the floor effect of retinal nerve fiber layer thickness affects measurements. Deep learning-based NR promises to improve glaucoma assessment.

Clinically Significant Nonperfusion Areas on Widefield OCT Angiography in Diabetic Retinopathy

Purpose

To investigate the distribution of clinically significant nonperfusion areas (NPAs) on widefield OCT angiography (OCTA) images in patients with diabetes.

Design

Prospective, cross-sectional, observational study.

Participants

One hundred and forty-four eyes of 114 patients with diabetes.

Methods

Nominal 20 × 23 mm OCTA images were obtained using a swept-source OCTA device (Xephilio OCT-S1), followed by the creation of en face images 20-mm (1614 pixels) in diameter centering on the fovea. The nonperfusion squares (NPSs) were defined as the 10 × 10 pixel squares without retinal vessels, and the ratio of eyes with the NPSs to all eyes in each square was referred to as the NPS ratio. The areas with probabilistic differences (APD) for proliferative diabetic retinopathy (PDR) and nonproliferative diabetic retinopathy (NPDR) (APD[PDR] and APD[NPDR]) were defined as sets of squares with higher NPS ratios in eyes with PDR and NPDR, respectively. The P ratio (NPSs within APD[PDR] but not APD[NPDR]/all NPSs) was also calculated.

Main Outcome Measures

The probabilistic distribution of the NPSs and the association with diabetic retinopathy (DR) severity.

Results

The NPSs developed randomly in eyes with mild and moderate NPDR and were more prevalent in the extramacular areas and the temporal quadrant in eyes with severe NPDR and PDR. The APD(PDR) was distributed mainly in the extramacular areas, sparing the areas around the vascular arcades and radially peripapillary capillaries. The APD(PDR) contained retinal neovascularization more frequently than the non-APD(PDR) (P = 0.023). The P ratio was higher in eyes with PDR than in those with NPDR (P < 0.001). The multivariate analysis designated the P ratio (odds ratio, 8.293 × 10⁷; 95% confidence interval, 6.529 × 10²-1.053 × 10¹³; P = 0.002) and the total NPSs (odds ratio, 1.002; 95% confidence interval, 1.001-1.003; P < 0.001) as independent risk factors of PDR. Most eyes with NPDR and 4-2-1 rule findings of DR severity had higher P ratios but not necessarily greater NPS numbers.

Conclusions

The APD(PDR) is uniquely distributed on widefield OCTA images, and the NPA location patterns are associated with DR severity, independent of the entire area of NPAs.

Financial Disclosures

Proprietary or commercial disclosure may be found after the references.

Ophthalmic imaging in diabetic retinopathy: A review

Retinal imaging has been a key tool in the diagnosis, evaluation, management and documentation of diabetic retinopathy (DR) and diabetic macular oedema (DMO) for many decades. Imaging technologies have rapidly evolved over the last few decades, yielding images with higher resolution and contrast with less time, effort and invasiveness. While many retinal imaging technologies provide detailed insight into retinal structure such as colour reflectance photography and optical coherence tomography (OCT), others such as fluorescein or OCT angiography and oximetry provide dynamic and functional information. Many other novel imaging technologies are in development and are poised to further enhance our evaluation of patients with DR.

Optical Coherence Tomography Angiography in Diabetic Patients: A Systematic Review

Background: Diabetic retinopathy (DR) is the leading cause of legal blindness in the working population in developed countries. Optical coherence tomography (OCT) angiography (OCTA) has risen as an essential tool in the diagnosis and control of diabetic patients, with and without DR, allowing visualisation of the retinal and choroidal microvasculature, their qualitative and quantitative changes, the progression of vascular disease, quantification of ischaemic areas, and the detection of preclinical changes. The aim of this article is to analyse the current applications of OCTA and provide an updated overview of them in the evaluation of DR. Methods: A systematic literature search was performed in PubMed and Embase, including the keywords “OCTA” OR “OCT angiography” OR “optical coherence tomography angiography” AND “diabetes” OR “diabetes mellitus” OR “diabetic retinopathy” OR “diabetic maculopathy” OR “diabetic macular oedema” OR “diabetic macular ischaemia”. Of the 1456 studies initially identified, 107 studies were screened after duplication, and those articles that did not meet the selection criteria were removed. Finally, after looking for missing data, we included 135 studies in this review. Results: We present the common and distinctive findings in the analysed papers after the literature search including the diagnostic use of OCTA in diabetes mellitus (DM) patients. We describe previous findings in retinal vascularization, including microaneurysms, foveal avascular zone (FAZ) changes in both size and morphology, changes in vascular perfusion, the appearance of retinal microvascular abnormalities or new vessels, and diabetic macular oedema (DME) and the use of deep learning technology applied to this disease. Conclusion: OCTA findings enable the diagnosis and follow-up of DM patients, including those with no detectable lesions with other devices. The evaluation of retinal and choroidal plexuses using OCTA is a fundamental tool for the diagnosis and prognosis of DR.

Microvasculature Features of Vogt-Koyanagi-Harada Disease Revealed by Widefield Swept-Source Optical Coherence Tomography Angiography

Background: Vogt-Koyanagi-Harada (VKH) disease is a multisystem autoimmune disorder which could induce bilateral panuveitis involving the posterior pole and peripheral fundus. Optical coherence tomography angiography (OCTA) provides several advantages over traditional fluorescence angiography for revealing pathological abnormalities of the retinal vasculature. Until recently, however, the OCTA field of view (FOV) was limited to 6 × 6 mm² scans.

Purpose: This study examined retinal vasculature and choriocapillaris abnormalities across multiple regions of the retina (15 × 9 mm² wide field, macular, peripapillary regions) among acute and convalescent VKH patients using a novel widefield swept-source OCTA (WSS-OCTA) device and assessed correlations between imaging features and best-corrected visual acuity (BCVA).

Methods: Twenty eyes of 13 VHK disease patients in the acute phase, 30 eyes of 17 patients in the convalescent phase, and 30 eyes of 15 healthy controls (HCs) were included in this study. Vascular length density (VLD) in superficial and deep vascular plexuses (SVP, DVP), vascular perfusion density (VPD) in SVP, DVP, and choriocapillaris (CC), and flow voids (FV) in CC were measured across multiple retinal regions via WSS-OCTA (PLEX Elite 9000, Carl Zeiss Meditec Inc., USA) using the 15 × 9 mm² scan pattern centered on the fovea and quantified by ImageJ.

Results: Compared to HCs, acute phase VKH patients exhibited significantly reduced SVP-VLD, SVP-VPD, and CC-VPD across multiple retinal regions (all p < 0.01). Notably, the FV area was more extensive in VKH patients, especially those in the acute phase (p < 0.01). These changes were reversed in the convalescent phase. Stepwise multiple linear regression analysis demonstrated that macular DVP-VLD and macular CC-VPD were the best predictive factors for BCVA in the acute and convalescent VKH groups.

Conclusion: The wider field of SS-OCAT provides more comprehensive and detailed images of the microvasculature abnormalities characterizing VKH disease. The quantifiable and layer-specific information from OCTA allows for the identification of sensitive and specific imaging markers for prognosis and treatment guidance, highlighting WSS-OCTA as a promising modality for the clinical management of VKH disease.

Enhanced Visualization of Retinal Microvasculature via Deep Learning on OCTA Image Quality

Purpose

To investigate the impact of denoising on the qualitative and quantitative parameters of optical coherence tomography angiography (OCTA) images of the optic nerve and macular area.

Methods

OCTA images of the optic nerve and macular area were obtained using a Canon-HS100 OCT device for 48 participants (48 eyes). Multiple image averaging (MIA) and denoising techniques were used to improve the quality of the OCTA images. The peak signal-to-noise ratio (PSNR) as an image quality parameter and vessel density (VD) as a quantitative parameter were obtained from single-scan, MIA, and denoised OCTA images. The parameters were compared, and the correlation was analyzed between different imaging protocols.

Results

In the optic nerve area, there were significant differences in the PSNR and VD in all measured regions between the three groups (P < 0.0001). The PSNR of the denoised group was significantly higher than that of the other two groups (P < 0.0001). The VD in the denoised group was significantly lower than that in the single-scan group in all measured regions (P < 0.0001). In the macular area, there were significant differences in the PSNR and VD in all measured regions among the three groups. The PSNR of the denoised group was significantly higher than that of the other two groups (P < 0.0001). The VD in the denoised group was significantly lower than that in the single-scan group in all measured regions. The VD around the optic nerve in the denoised group was correlated with that in the single-scan group (R = 0.9403, P < 0.0001), but the VD in the MIA group was not correlated with that in the single-scan group (R = 0.2505, P = 0.2076). The VD around the fovea in the denoised and MIA images was correlated with that in the single-scan group (R = 0.7377, P < 0.0001; R = 0.7005, P = 0.0004, respectively).

Conclusion

Denoising could provide an easy and quick way to improve image quality parameters, such as PSNR. It shows great potential in improving the sensitivity of OCTA images as retinal disease markers.