Multimodal dataset using OCTA and fundus images for the study of diabetic retinopathy

This article presents a Multimodal database consisting of 222 images of 76 people wherein 111 are OCTA images and 111 are color fundus images taken at the Natasha Eye Care and Research Institute of Pune Maharashtra, India. Nonmydriatic fundus images were acquired using a confocal SLO widefield fundus imaging Eidon machine. Nonmydriatic OCTA images were acquired using the Optovue Avanti Edition machine Initially, the clinical approach described in this article was used to obtain the retinal images. Following that, the dataset was categorized by two experienced eye specialists. To identify instances of Non-Proliferative Diabetic Retinopathy (NPDR) with their various stages, medical professionals and scholars can use this data. Research scholars and ophthalmologists can utilize the data created to develop the initial stages of automated identification techniques for diabetic retinopathy (DR).

Instructions for accessing these data: Access will be given on prior permission specifying the use of the dataset and with proper citation and DOI

Value of the Data
The multimodal dataset serves as a valuable resource for both the medical scientific community and informatics researchers.It plays a crucial role in several aspects: • This multimodal data set is particularly helpful for training and conducting research by novice experts who are interested in the examination of DR at the early stage.• As this is a multimodal dataset with Fundus and OCTA images, by comparing this with normal images, the data set can be used to study retinal anatomy and its physiological variations within diabetic and nondiabetic patients.• Encouraging the creation of mathematical frameworks for automated medical diagnosis, particularly for the use of AI tools for detecting diabetic retinopathy (DR) early.• Supporting the training of future physicians in identifying and analyzing various stages of DR using fundus and OCTA images [1 , 2] .

Data Description
Retinal or fundus images play a crucial role in the identification of various eye-related disorders.These images capture essential structures within the eye, including blood vessels, the optic disc, and the macula.The information derived from these structures in retinal images is instrumental in diagnosing and treating a range of retinal conditions [3][4][5] .
The Early Treatment Diabetic Retinopathy Study (ETDRS) classifies diabetic retinopathy into two main categories: Non-Proliferative Diabetic Retinopathy (NPDR) and Proliferative Diabetic Retinopathy (PDR).NPDR represents the early stage of the disease and further divides into mild, moderate, severe, and very severe.On the other hand, PDR is a more advanced form and is categorized into early and advanced stages.Medical experts have categorized the dataset into three distinct groups based on the pathological conditions associated with diabetic retinopathy.Consequently, the dataset is organized into three disease categories, including No DR signs, Mild (or early) NPDR, and Moderate NPDR For detailed information on the criteria used for classifying fundus images, please refer to [5] .The new dataset comprises 222 fundus and OCTA images of 76 persons where some patients have diabetes, and some patients don't have diabetes with corresponding labels and classification status available in the "Annotations of the classifications.xlsx"file.Fig. 1 shows OCTA and Fundus images with different stages of DR.

Experimental Design, Materials and Methods
In this study total of 222 OCTA and Fundus images were obtained from individuals over the age of 30 years and above through a specific process.Cataract patients of similar age groups wherein some of them have diabetes and some of them don't have diabetes also some patients who have had diabetes for a very long time are considered.Post Operatively on Day-15 when individuals come for the second follow-up their OCTA and Fundus Scans are taken at Natasha Eye Care and Research Institute Pune.The described procedure is outlined as follows:

Limitations
Describing the research and obtaining patients' consent poses a significant challenge [ 7,8 ].Additionally, performing an OCTA scan requires expert guidance because it necessitates more patient cooperation to achieve optimal clarity.This study is limited to 222 multimodal images, which includes 111 OCTA images and 111 Fundus images of 76 patients.

Ethics Statement
The study was approved by the Symbiosis Institutional Ethics Committee (BHR) SIU with approval number SIU/IEC/583.Every patient provided informed consent for both treatment and examination taken by the Department of Ophthalmology at Natasha Eye Care and Research Institute.Data is protected by a data usage agreement.The patients' identities remain undisclosed, and the confidentiality of their medical conditions is rigorously maintained.

Fig. 1 a
Fig. 1 shows OCTA and Fundus images with different stages of DR.Fig. 1 a provides clear OCTA and Fundus images of the same person with no signs of DR.Microaneurysms, or any other lesions are absent in this case.Fig. 1 b provides OCTA and Fundus images of the same person with Mild signs of DR.Very few microaneurysms are visible in both the modalities in case of Mild DR.Fig. 1 c provides OCTA and Fundus images of the same person with Moderate signs of DR.In this type of DR, microaneurysms, along with retinal haemorrhages or exudates are visible.

Fig. 1 .
Fig. 1. a OCTA and Fundus images with No signs of DR.Fig. 1b OCTA and Fundus images with Mild signs of DR.Fig. 1c OCTA and Fundus images with Moderate signs of DR.
Fundus images with dimensions 3680 × 3288, 96 dpi, and jpeg 2. OCTA (8 × 8 mm) images with dimensions 1596 × 990, 96 dpi and jpeg 3. Annotations of the classifications are provided in an EXCEL file (.xlsx).Data collection This dataset contains OCTA and Fundus images.Nonmydriatic fundus images were acquired using a confocal SLO widefield fundus imaging Eidon machine and Nonmydriatic OCTA images were acquired using the Optovue Avanti Edition machine.OCTA and Fundus images are classified into three categories NO DR signs, Mild DR, and Moderate DR.

Table 2
Categorization of the images within the database.
1. Patients undergo an interview, during which their information is documented in a patient file, including a checklist to document details regarding the patient's age and diabetes history.Additional health conditions or concurrent illnesses such as hypertension, ischemic heart disease (IHD), or the type of medication they are using, which may include oral pills, insulin, or a combination of both.2. For advanced cases, systemic factors play a crucial role, including conditions like nephropathy, neuropathy, retinopathy, cardiomyopathy, vasculopathy, gastroparesis, and skin issues.3. OCTA Images were obtained as 8 × 8 mm2 in size, centered on the macula, and images of superficial capillary plexus (SCP), Deep capillary plexus (DCP), and full thickness retinal slab are taken from Optovue Avanti Edition OCTA Machine with the Specifications : Machine Name: RTVue XR 100 Avanti Edition, FOV 32 o (H) x 22 o (V), Monochrome CCD Camera: WVGA 1/3 o CCD Format, NIR Illumination: 735 nm LED. 4. Fundus Scans are taken using Eidon machine with the Specifications as Centervue Eidon UWF Confocal SLO, Pupil size-2.5mm,110 o wide view scanning technique in automatic mode & 150 o in manual mode. 5.The determination of diabetic retinopathy stage is determined by a retina specialist.During the assessment of acquired retinographies, pictures that are out of focus, contain artifacts, or exhibit low image quality due to issues like low signal strength are excluded.6.Additionally, retinographies from patients with any other retina treatment, patients with macular edema [6] , and macular drusens, Proliferative Diabetic Retinopathy Patients, postlaser patients, and PVD (Posterior Vitreous Detachments) patients are not incorporated.