Does the microaneurysm have to be distinguished from a similar-sized dot hemorrhage?

To the Editor: China now has the largest number of patients with diabetes mellitus (DM) worldwide, making diabetic retinopathy (DR) the leading cause of blindness in the Chinese working-age population.[1] Fundus screening for patients with DM is important so that ophthalmologists treat retinopathy in a timely manner and delay the development of vision loss. However, we found some problems in the clinical practice of fundus screening. Microaneurysm (MA) and small dot intraretinal hemorrhage (HEM), although described precisely by the Early Treatment Diabetic Retinopathy Study (ETDRS) group,[2] are red and round spots on dilated ophthalmoscopy and cannot be easily differentiated due to incompletely transparent dioptric media of patients unless fundus fluorescein angiography (FFA) is performed. Questions always come up: Do we have to differentiate these two lesions when grading the severity of DR? In this study, we investigated the ability of Chinese ophthalmologists to differentiate MA from dot HEM on fundus images of patients with DR and discussed the possibility of not distinguishing them. A questionnaire-based cross-sectional survey was conducted. The questionnaire was distributed online through the WeChat groups of ophthalmologists from December 15 to 18, 2017. The questionnaire had two parts, personal basic information and choice questions. Personal basic information included professional titles, years of working in ophthalmology, specializing in retinal diseases or not, and hospital level [Supplementary Table 1, https://links.lww.com/CM9/C27]. The question part consisted of 20 multiple-choice questions. We chose a color fundus photography of the left eye of a severe non-proliferative diabetic retinopathy (NPDR) patient [Supplementary Figure 1, https://links.lww.com/CM9/C27] taken with a Topcon TRC-50DX retinal camera (Topcon Optical Company, Tokyo, Japan). Moreover, in each question, a red and round spot was indicated in this picture, and the respondents chose whether it was an MA or a dot HEM, as shown in Supplementary Figure 2, https://links.lww.com/CM9/C27. This patient had a transparent left lens and the photo was taken after mydriasis. The picture we chose has relatively high clarity, although not the clearest, better than most non-mydriatic fundus images used in DR screening. Among the 20 small spots whose character was confirmed by FFA in advance, there were 11 MAs and 9 dot HEMs. A right choice could get 1 point, and the total score was 20 points. The study population was limited to ophthalmologists from different provinces of China with different seniority and subspecialties. SSPS statistical software (SPSS for Windows, version 26.0; SPSS, Inc., Chicago, IL, USA) was used for statistical analysis. A two-sided P value <0.05 was considered statistically significant. We obtained 2031 valid questionnaires from ophthalmologists from all over China. The respondents were from provinces and municipalities in China, with the highest number in Sichuan (175 ophthalmologists) and Beijing (148 ophthalmologists), accounting for 8.62% and 7.29%, respectively [Supplementary Figure 3, https://links.lww.com/CM9/C27]. The 396 respondents who are majoring in the medical retina would be considered the most qualified group to treat DR problems. The accuracy was 33.1−77.0% for MAs and 27.0−72.0% for HEMs [Supplementary Table 2, https://links.lww.com/CM9/C27]. Among these 2031 respondents, the score ranged from 3 points to 17 points with an average of 10.88 ± 2.09 points, and the total accuracy rate was approximately 54% and was similar to the result of random choice. Comparisons among the different groups are shown in Supplementary Figure 4, https://links.lww.com/CM9/C27. Among ophthalmologists with different professional titles, professors scored the highest average score of 11.00 ± 1.86 points, but no statistically significant differences were found among groups with different titles (F = 0.786, P = 0.501). We also found no statistically significant differences between ophthalmologists with different years of working (F = 0.890, P = 0.469). The scores of ophthalmologists working in tertiary hospitals were higher than those in secondary hospitals (P = 0.006). Ophthalmologists, the most qualified group specialized in retinal diseases, had higher points than those specialized in others (P = 0.001). They had the highest average points than other groups, but 11.06 ± 2.03 remained a non-satisfactory score to match their seniority. This result indicates that it is difficult to differentiate MA from similar-sized dot HEM with naked eyes on a dilated fundus image, even for highly experienced retinal doctors. In a practical setting, it is difficult for ophthalmologists to see all the little details of the fundus, whether a dilated ophthalmoscopy is performed or reading the fundus images. Retinal MA is a hallmark of retinal microvascular disease in diabetic patients and is the first lesion found on fundus examination, which involves many causal or contributing histopathological factors, such as damage to the vascular basement membrane, changes in intraluminal rheology, and pericyte cell death. It has been speculated that MAs may result from asymmetrical dilatations of the capillary wall, where it is weakened or damaged, following the loss of the supporting pericytes and localized increases in hydrostatic pressure. HEM is an associated change in microvascular function caused by the rupture of the dilated capillary wall.[3] The clinical significance of MAs and similar-sized dot HEM differs from a pathological perspective. They also represent different meanings. Although MA and HEM are different pathological changes, they occur successively in the early stage of DR. The distinction between the two does not affect the development of treatment plans or follow-up time. According to our study, MA and dot HEM are not easily distinguishable by ophthalmoscopy or ordinary fundus photography [Supplementary Figure 5, https://links.lww.com/CM9/C27]. It is difficult for novice ophthalmologists and retinal disease professionals to accurately differentiate MA from a dot HEM. Retinal disease professionals often refer to FFA results to comprehensively determine the grading and treatment plan for DR. However, without the assistance of FFA and optically correlated tomography angiography (OCTA), counting the number of MAs and dot HEMs for DR grading can cause difficulties for general ophthalmologists. The most widely used Chinese DR classification (2014)[2] refers to the 2003 International Diabetes Retinopathy Severity Scale.[4] Its definition of "4" in the "4–2–1" rule is "more than 20 intraretinal HEMs in each of the four quadrants," which requires us to distinguish MA from dot HEM when counting. It caused difficulties in clinical practice. The Diabetic Retinopathy Preferred Practice Pattern (PPP) issued by the American Academy of Ophthalmology in 2016[5,6] provided a different definition of the "4–2–1" rule from the 2003 international version, weakening the distinction between the two in grading and assembling them together. The definition of "4" in the "4–2–1" rule as "severe intraretinal HEMs and MAs in each of the four quadrants" indicates that PPP does not need to distinguish MA from a dot HEM in DR grading deliberately. This study aimed to show that it is not necessary to distinguish whether a lesion is an MA or a dot HEM when applying DR grading. It cannot be clearly distinguished by ophthalmoscopy or photographs. The absence of a forcible distinction between the two makes DR screening more convenient and feasible for many novice ophthalmologists, even internal medicine doctors. DR screening by artificial intelligence (AI) is based on the principle of deep learning, with doctors first marking typical lesions and disease grades, and then performing machine learning and training.[7] However, through our investigation, it has been found that even experienced retinal doctors sometimes find it difficult to distinguish MA from similar-sized dot HEM, not to mention that machine deep learning has the ability to distinguish them accurately. Therefore, not emphasizing the distinction between the two in grading can also make AI algorithms easier. Although clear discrimination of MA and HEM is ideal for the detection of DR, the reality is that most ophthalmologists have trouble distinguishing them. The detection and research of DR will continue to progress and we want to mention and discuss this little puzzle that we meet in clinical practice. We hope that this paper will provide insight to the readers. Funding The present study was supported in part by the Sichuan Provincial Science and Technology Support Project (Nos. 2022YFS0192 and 2021YFS0213).