Ability of MultiColor scanning laser ophthalmoscope to detect non-glaucomatous retinal nerve fiber layer defects in eyes with retinal diseases

Colour fusion effect on deep learning classification of uveal melanoma

BACKGROUND

Reliable differentiation of uveal melanoma and choroidal nevi is crucial to guide appropriate treatment, preventing unnecessary procedures for benign lesions and ensuring timely treatment for potentially malignant cases. The purpose of this study is to validate deep learning classification of uveal melanoma and choroidal nevi, and to evaluate the effect of colour fusion options on the classification performance.

METHODS

A total of 798 ultra-widefield retinal images of 438 patients were included in this retrospective study, comprising 157 patients diagnosed with UM and 281 patients diagnosed with choroidal naevus. Colour fusion options, including early fusion, intermediate fusion and late fusion, were tested for deep learning image classification with a convolutional neural network (CNN). F1-score, accuracy and the area under the curve (AUC) of a receiver operating characteristic (ROC) were used to evaluate the classification performance.

RESULTS

Colour fusion options were observed to affect the deep learning performance significantly. For single-colour learning, the red colour image was observed to have superior performance compared to green and blue channels. For multi-colour learning, the intermediate fusion is better than early and late fusion options.

CONCLUSION

Deep learning is a promising approach for automated classification of uveal melanoma and choroidal nevi. Colour fusion options can significantly affect the classification performance.

Clinical application of multicolor scanning laser ophthalmology in diagnosis and grading of central retinal artery occlusion

Purpose

To characterize features of central retinal artery occlusion (CRAO) using multicolor (MC) imaging and to assess the differences in CRAO grading between color fundus photography (CFP) and MC image qualitatively and quantitatively.

Methods

We conducted a prospective, cross-sectional study in the Department of Ophthalmology of Renmin Hospital of Wuhan University. In total, 86 acute CRAO patients were included. Spectral-domain optical coherence tomography (SD-OCT), CFP, and MC examinations were taken at baseline. Based on the findings of these three examinations, CRAO was divided into three grades (incomplete, subtotal, and total). Based on OCT grading criteria, we qualitatively compared the ability of grading CRAO by CFP and MC. CRAO patient's visual acuity (VA) was obtained from the initial visit. The retinal thickness was measured by SD-OCT. Superficial capillary plexus (SCP) and deep capillary plexus (DCP) were obtained from optical coherence tomography angiography (OCTA) examinations. Quantitative data were compared across the three acute CRAO subgroups and against three examination findings.

Results

MC image had significantly higher power of acute CRAO detection than CFP (P = 0.03). In the same group of CRAO patients, there was no significant difference in VA when comparing OCT with the MC grading system or with the CFP grading system (all P > 0.05). Significant differences in VA were found between the three CRAO subgroups only under MC grading (P = 0.016). In incomplete CRAO patients, significant differences were found in central fovea thickness (CFT) when comparing OCT with the CFP grading system (P = 0.019). In the same group of CRAO patients, there was no significant difference in retinal thickness when comparing OCT with the MC grading system (All P > 0.05). Significance differences in CFT (P < 0.001), innermost retinal layer (IMRL; P < 0.01), middle retinal layer (MRL; P < 0.001), and outer retinal layer (ORL; P = 0.021) were found between the three CRAO subgroups by MC grading. Vessel density of SCP showed a statistically increased as the severity of three CRAO subgroups (P = 0.03), whereas DCP did not have significant differences (P = 0.745). Comparisons were made between the OCT grading method and the MC and CFP grading methods; there is no significant difference in vessel density of SCP and DCP (All P > 0.05).

Conclusion

The images obtained by MC are superior to those obtained by CFP in CRAO grading, retinal thickness, and vessel density measurement. MC imaging may be more capable of CRAO grading than OCT. We recommend MC imaging to determine CRAO severity to guide disease treatment and predict visual prognosis.

Color Fusion Effect on Deep Learning Classification of Uveal Melanoma

Background

Reliable differentiation of uveal melanoma and choroidal nevi is crucial to guide appropriate treatment, preventing unnecessary procedures for benign lesions and ensuring timely treatment for potentially malignant cases. The purpose of this study is to validate deep learning classification of uveal melanoma and choroidal nevi, and to evaluate the effect of color fusion options on the classification performance.

Methods

A total of 798 ultra-widefield retinal images of 438 patients were included in this retrospective study, comprising 157 patients diagnosed with UM and 281 patients diagnosed with choroidal nevus. Color fusion options, including early fusion, intermediate fusion and late fusion, were tested for deep learning image classification with a convolutional neural network (CNN). Specificity, sensitivity, F1-score, accuracy, and the area under the curve (AUC) of a receiver operating characteristic (ROC) were used to evaluate the classification performance. The saliency map visualization technique was used to understand the areas in the image that had the most influence on classification decisions of the CNN.

Results

Color fusion options were observed to affect the deep learning performance significantly. For single-color learning, the red color image was observed to have superior performance compared to green and blue channels. For multi-color learning, the intermediate fusion is better than early and late fusion options.

Conclusion

Deep learning is a promising approach for automated classification of uveal melanoma and choroidal nevi, and color fusion options can significantly affect the classification performance.

Sex determination using color fundus parameters in older adults of Kumejima population study

PURPOSE

Deep learning artificial intelligence can determine the sex using only fundus photographs. However, the factors used by deep learning to determine the sex are not visible. Therefore, the purpose of the study was to determine whether the sex of an older individual can be determined by regression analysis of their color fundus photographs (CFPs).

METHODS

Forty-two parameters were analyzed by regression analysis using 1653 CFPs of normal subjects in the Kumajima study. The parameters included the mean values of red, green, and blue intensities; the tessellation fundus index; the optic disc ovality ratio; the papillomacular angle; and the retinal vessel angles. Finally, the L2 regularized binomial logistic regression was used to predict the sex using all the parameters, and the diagnostic ability was assessed through the leave-one-cross-validation.

RESULTS

The mean age of the 838 men and 815 women were 52.8 and 54.0 years, respectively. The ovality ratio and retinal artery angles in women were significantly smaller than that in men. The green intensity at all locations for the women were significantly higher than that of men (P < 0.001). The discrimination accuracy rate assessed by the area-under-the-curve was 80.4%.

CONCLUSIONS

Our methods can determine the sex from the CFPs of the adult with an accuracy of 80.4%. The ovality ratio, retinal vessel angles, tessellation, and the green intensities of the fundus are important factors to identify the sex in individuals over 40 years old.

Clinical application of multicolor imaging in Leber hereditary optic neuropathy

Purpose

To characterize features of retinal never fiber in Leber Hereditary Optic Neuropathy (LHON) using multicolor (MC) imaging and color fundus photography (CFP).

Methods

Ninety-two eyes of patients with LHON underwent MC imaging, optic disc spectral domain optical coherence tomography (SD-OCT), and CFP. Two independent observers graded RNFL visibility scores and two other experts determined never fiber bundle defects from four-quadrant readings. CFP, standard MC, infrared reflectance (IR), green reflectance (GR), blue reflectance (BR), and green-blue-enhanced (BG) imaging were compared.

Results

Agreement on never fiber bundle defects was substantial for CFP, standard MC, GR, BR, and BG images relative to IR. It was shown that BR (2.71 ± 0.55) had the best mean RNFL visibility score, BG (2.69 ± 0.52), GR (2.69 ± 0.56), standard MC (2.04 ± 0.79), CFP (1.80 ± 0.82), and IR (0.45 ± 0.59) followed. Agreement on temporal area defects was relatively improved. Youden's indices for CFP (78.21%), standard MC (84.48%), GR (90.92%), BR (89.64%), and BG (90.99%) indicated good detection of defects in the papillomacular bundle (PMB)/ high suspicion of patients with LHON, particularly for BG and GR. According to the proportion of never fiber bundle defects, standard MC, GR, BR, and BG can roughly determine the LHON clinical stage, especially in subacute and chronic stages, and standard MC is superior for patients with LHON of all stages. The stage judged by MC was consistent with the course inferred by pRNFL thickness.

Conclusion

As an adjunct to SD-OCT, the MC image, particularly the GR and BG can delineate RNFL more effectively than CFP. The MC image may be a useful adjunct to OCT for detecting or monitoring never fiber bundle defects, providing inexpensive and rapid methods that can quickly identify patients with high suspicion of LHON.

Multicolor Imaging for Detection of Retinal Nerve Fiber Layer Defect in Myopic Eyes With Glaucoma

PURPOSE

To investigate the diagnostic performance of multicolor imaging for detection of retinal nerve fiber layer (RNFL) defects in myopic eyes with glaucoma.

DESIGN

Comparative diagnostic analysis.

METHODS

A total of 150 myopic eyes of 138 patients (< -1.00 diopter or axial length >24 mm) underwent multicolor imaging by Spectralis (Heidelberg Engineering GmbH) optical coherence tomography (OCT). Two independent observers graded the RNFL visibility score and determined the location of the RNFL defect on each image. Readings were taken from the superior and inferior areas. The diagnostic performances of multicolor, green-reflectance, and blue-reflectance images for detection of the RNFL defect were compared with those of conventional fundus and red-free RNFL photographs.

RESULTS

The interobserver agreement on the RNFL visibility score was substantial for fundus and RNFL photographs and multicolor and green-reflectance images and was perfect for blue-reflectance images. The interobserver agreement on the location of the RNFL defect was better when using multicolor imaging than when using conventional photography. The green-reflectance image (2.89 ± 0.31) showed the highest RNFL visibility score, followed by blue-reflectance (2.88 ± 0.32) and multicolor (2.57 ± 0.63) images. The sensitivity of multicolor imaging (94.9%, 93.2%, and 91.5% for green-reflectance, blue-reflectance, and multicolor images, respectively) was significantly higher than that of fundus photography (83.1%) for detection of superotemporal RNFL defects (all P < .05). It was significantly higher for inferotemporal defects than those of fundus and RNFL photographs (all P < .05).

CONCLUSIONS

Multicolor imaging can visualize the glaucomatous RNFL defects in myopic eyes that are indistinguishable on conventional photography.

Blue Widefield Images of Scanning Laser Ophthalmoscope Can Detect Retinal Ischemic Areas in Eyes With Diabetic Retinopathy

PURPOSE

To determine whether the hyporeflective areas in the blue images obtained by widefield scanning laser ophthalmoscopy (SLO) correspond to the non-perfused areas (NPAs) in the fluorescein angiograms (FA) in eyes with diabetic retinopathy (DR).

DESIGN

Retrospective observational case series.

METHODS

Ninety patients with diabetes mellitus (DM) were studied. All had undergone multicolor widefield SLO imaging. The NPAs in the FA images and hyporeflective areas in the blue widefield SLO images were examined. The morphology of the retina was determined by optical coherence tomography.

RESULTS

Hyporeflective areas in the blue SLO images were found with a rate of 76.6% in eyes with proliferative DR eyes. In a comparison of the hyporeflective areas of the blue SLO images to the non-perfused areas in the FA images, the appearance and the correspondence in the locations of these two types of images were found, and the rate was highly concordant with a Cohen's kappa value of 0.675.

CONCLUSIONS

The high concordance between the hyporeflective areas in the widefield blue SLO and the NPAs in the FA indicates that widefield blue SLO can be used to identify ischemic retinal areas in eyes with DR without the intravenous injection of any dye.

MORE EFFECTIVE SCREENING FOR EPIRETINAL MEMBRANES WITH MULTICOLOR SCANNING LASER OPHTHALMOSCOPE THAN WITH COLOR FUNDUS PHOTOGRAPHS

PURPOSE

To assess the ability of the multicolor scanning laser ophthalmoscope (MC-SLO) to screen for epiretinal membranes (ERMs).

METHODS

A retrospective cross-sectional study of 35 eyes of 32 patients with an ERM detected by optical coherence tomography and 46 eyes of 23 healthy volunteers. The detection of the ERM was graded into three visibility scores-1, not visible, 2, barely visible, and 3, clearly visible-by retina specialists or by ophthalmology residents. The sensitivity and specificity of the detection with the merged image of the MC-SLO or color fundus photographs (CFPs) were calculated.

RESULTS

The sensitivity for ERM detection in the MC-SLO and CFP were 91.4% and 65.7% by specialists and 97.1% and 60.0% by residents. The specificity for both devices was 100% by specialists and residents. The visibility score for the MC-SLO images were significantly higher than that for the CFP by both specialists and residents. In addition, the visibility score for the MC-SLO determined by residents was significantly higher than that for the CFP by specialists.

CONCLUSION

The detection of an ERM is better with the MC-SLO images than with CFP. Furthermore, the ERM detection in the MC-SLO images by residents was comparable to that by specialists.

Recent Advances and Clinical Application of Color Scanning Laser Ophthalmoscope

Scanning laser ophthalmoscopes (SLOs) have been available since the early 1990s, but they were not commonly used because their advantages were not enough to replace conventional color fundus photography. In recent years, color SLOs have improved significantly, and the colored SLO images are obtained by combining multiple SLO images taken by lasers of different wavelengths. A combination of these images of different lasers can create an image that is close to that of the real ocular fundus. One advantage of the advanced SLOs is that they can obtain images with a wider view of the ocular fundus while maintaining a high resolution even through non-dilated eyes. The current SLOs are superior to the conventional fundus photography in their ability to image abnormal alterations of the retina and choroid. Thus, the purpose of this review was to present the characteristics of the current color SLOs and to show how that can help in the diagnosis and the following of changes after treatments. To accomplish these goals, we will present our findings in patients with different types of retinochoroidal disorders.

Multicolor Scanning Laser Ophthalmoscopy Strengthens Surgeons' Preoperative Decision-Making and Intraoperative Performance on Epiretinal Membrane

Purpose

To determine whether multicolor scanning laser ophthalmoscopy (MC-SLO) was better than color fundus photography (CFP) to enhance residents and specialists’ preoperative decision-making and intraoperative performance on the epiretinal membrane (ERM).

Methods

Consecutive patients with idiopathic ERM were recruited prospectively. All the patients underwent MC-SLO and CFP imagings and were randomized into MC-SLO (n = 20) and CFP (n = 20) groups. Preoperatively, residents and specialists were required to have ERM delineation and select an optimal location for initial ERM peeling independently, based on the MC-SLO (MC-SLO group) or CFP (CFP group) images. Intraoperative optical coherence tomography (iOCT) was introduced to evaluate the accuracy.

Results

Preoperatively, residents and specialists acted more effectively in ERM delineation and selection of initial grasping location in the MC-SLO group (both P < 0.001). In the MC-SLO group, higher resident–specialist agreements were achieved in ERM delineation (P = 0.002) and selection of initial grasping location (P = 0.035). The iOCT revealed greater interobserver (iOCT–resident and iOCT–specialist) agreements of ERM delineation in MC-SLO group (P < 0.001 and = 0.027, respectively). Surgeons acted more effectively on completely peeling the ERM in the MC-SLO group (P < 0.001).

Conclusions

MC-SLO provided a better visual reference for residents and specialists in ERM delineation and the selection of an initial grasping location for the surgery, compared with CFP.

Translational Relevance

MC-SLO is able to help surgeons achieve better intraoperative performance and shorten the learning process for residents.

Sex judgment using color fundus parameters in elementary school students

PURPOSES

Recently, artificial intelligence has been used to determine sex using fundus photographs alone. We had earlier reported that sex can be distinguished using known factors obtained from color fundus photography (CFP) in adult eyes. However, it is not clear when the sex difference in fundus parameters begins. Therefore, we conducted this study to investigate sex determination based on fundus parameters using binominal logistic regression in elementary school students.

METHODS

This prospective observational cross-sectional study was conducted on 119 right eyes of elementary school students (aged 8 or 9 years, 59 boys and 60 girls). Through CFP, the tessellation fundus index was calculated as R/(R + G + B) using the mean value of red-green-blue intensity in the eight locations around the optic disc. Optic disc ovality ratio, papillomacular angle, retinal artery trajectory, and retinal vessel were quantified based on our earlier reports. Regularized binomial logistic regression was applied to these variables to select the decisive factors. Furthermore, its discriminative performance was evaluated using the leave-one-out cross-validation method. Sex difference in the parameters was assessed using the Mann-Whitney U test.

RESULTS

The optimal model yielded by the Ridge binomial logistic regression suggested that the ovality ratio of girls was significantly smaller, whereas their nasal green and blue intensities were significantly higher, than those of boys. Using this approach, the area under the receiver-operating characteristic curve was 63.2%.

CONCLUSIONS

Although sex can be distinguished using CFP even in elementary school students, the discrimination accuracy was relatively low. Some sex difference in the ocular fundus may begin after the age of 10 years.

Factors in Color Fundus Photographs That Can Be Used by Humans to Determine Sex of Individuals

Purpose

Artificial intelligence (AI) can identify the sex of an individual from color fundus photographs (CFPs). However, the mechanism(s) involved in this identification has not been determined. This study was conducted to determine the information in CFPs that can be used to determine the sex of an individual.

Methods

Prospective observational cross-sectional study of 112 eyes of 112 healthy volunteers. The following characteristics of CFPs were analyzed: the color of peripapillary area expressed by the mean values of red, green, and blue intensities, and the tessellation expressed by the tessellation fundus index (TFI). The optic disc ovality ratio, papillomacular angle, retinal artery trajectory, and retinal vessel angles were also quantified. Their differences between the sexes were assessed by Mann-Whitney U tests. Regularized binomial logistic regression was used to select the decisive factors. In addition, its discriminative performance was evaluated through the leave-one-out cross validation.

Results

The mean age of 76 men and 36 women was 25.8 years. The regularized binomial logistic regression delivered the optimal model for sex selected variables of peripapillary temporal green and blue intensities, temporal TFI, supratemporal TFI, optic disc ovality ratio, artery trajectory, and supratemporal retinal artery angle. With this approach, the discrimination accuracy rate was 77.9%.

Conclusions

Human-assessed characteristics of CFPs are useful in investigating the new theme proposed by AI, the sex of an individual.

Translational Relevance

This is the first report to approach the thinking process of AI by humans and can be a new approach to medical AI research.