Ultra-wide-field fundus photography compared to ophthalmoscopy in diagnosing and classifying major retinal diseases

Although indirect ophthalmoscopy, particularly associated to scleral depression, and posterior biomicroscopy still remain the gold standard for fundus examination¹⁹, over the last ten years the use of fundus photography system has extensively increased^11,12,20, as an adjunct to clinical evaluation^7,21,22 or as a screening tool for many retinal diseases, in particular diabetic retinopathy^4,5,6,10, peripheral retinal lesions^8,9 and other retinal disorders^16,23,24, also by means of deep learning technologies^3,6,25,26. This is due to the progressive improvement of retinal fundus cameras, involving both the extent of retinal field evaluated and the quality of images: from the earliest cameras encompassing a retinal field of 20–30° in a single image ¹², we have now systems capable of acquiring real color fundus images covering up to 200° of retinal field with a pixel definition of 6604 × 4274 (Clarus™), or providing a two-tone fundus image of 200° of field with a 3900 × 3072 pixel definition (Optos®)^4,12,14. The COVID-19 pandemic has exacerbated this trend due to the need for reducing close contacts and the burden on health-care systems^1,2,3: besides safety, eye tele-screening reduces the time of the examination both in adults and in children³.

Recent studies evaluated UWF imaging for individual disorders, most of which performed on diabetic retinopathy, peripheral retinal lesions or other retinal disorders. Some reports have shown that UWF fundus imaging is an effective useful tool for the assessment of diabetic retinopathy^4,5,12, focusing on the higher accuracy of Clarus™ in detecting microaneurysm and retinal hemorrhages: by providing real color images and reducing lids and lashes artifacts, it allows a slightly more precise staging of diabetic retinopathy and maculopathy than Optos® camera⁴. Conversely, studies assessing the effectiveness of UWF imaging in detecting peripheral retinal lesions, such as retinal degenerations, retinal breaks, rhegmatogenous retinal detachment, have been conducted primarily with Optos®: despite considering UWF imaging a useful adjunct to medical evaluation, some authors do not agree it may represent substitute of clinical ophthalmoscopy due to the possible missing of some peripheral retinal lesions^7,21. Other authors observed consistent findings between clinical and UWF imaging examination, and consider the two methods interchangeable^8,19. Moreover, even if the effective fields of views between Clarus™ and Optos® seem different depending on the specific retinal quadrant¹⁴, a recent study found a similar ability to detect treatment-requiring retinal breaks between the two systems⁹. UWF imaging appeared to be useful also for screening of ocular Toxoplasmosis²³ and even superior to dilated fundus examination for the screening of sickle cell retinopathy, because of the higher accuracy in detecting capillary occlusion or anastomosis¹⁶, and in inherited retinal dystrophies, providing previously unavailable information about retinal periphery²⁴.

Our study was planned to compare retinal UWF imaging versus clinical evaluation and to validate the technique of clinical analysis of retinal UWF imaging, obtained by means of Clarus™, performed without a deep learning system. We enrolled patients evaluated in daily clinics for presumed major retinal disorders in a period of about two years, without differentiating for individual disorder, thus collecting a very large unselected population (7024 eyes).

Our results demonstrated that UWF imaging, by means of Clarus™ fundus color images, is comparable to ophthalmoscopic examination performed with indirect ophthalmoscopy plus macular biomicroscopy, when the macula appears to be involved. For major retinal diseases these evaluation systems have similar sensitivity and specificity (almost 100%) and no statistically significant difference was found in fundus assessment between the two procedures (k = 0.997, 95%CI = 0.996–0.999). These findings differ from some previous reports^7,9,21, but it must be underlined that the other study populations were small and mostly limited to peripheral retinal degenerations^7,9,21. Moreover, these last reports used Optos® system^7,21, whose limitations, mainly concerning true color, have already been reported⁴. On the other hand, our results are consistent with some recent studies^{4,5,8,16,19,23}, even if these studies were also performed on small populations, for selected diagnosis (diabetic retinopathy^4,5, peripheral lesions⁸, rhegmatogenous retinal detachment¹⁹, ocular Toxoplasmosis²³, sickle cell retinopathy¹⁶). In accordance with another analysis ¹⁴, the missing diagnosis we reported were very limited in number and due to lesions located in the extreme upper temporal periphery and statistically no significant. This seem to be related to the specific examined quadrant by the limitation induced by the patient’s nose when trying to capture images of the extreme temporal periphery, since most of the missed lesions were located in the upper temporal periphery (4 cases)⁹. The classification and diagnosis achieved with UWF imaging allowed to address the patient to the correct management, such as adequate follow-up or, when needed, directing the patient to a specific clinical care pathway for the treatment of its specific retinal disorder.

The relevance of this clinical study also lies in analyzing UWF fundus photographs of such a broad population, both in terms of numbers and disorders. Moreover, the use of Clarus™ system, providing real color fundus imaging, allows realistic and accurate evaluation, comparable to the clinical one with ophthalmoscopy, except for some rare cases. However, according to patients’ symptoms and clinical features and/or to UWF photos characteristics, the examination also with indirect ophthalmoscopy may be necessary for an accurate diagnosis. In fact, at present, the approach to patients affected by retinal diseases should always be multimodal, and UWF fundus photo might be the first diagnostic modality in this approach, followed by the other, eventually clinically-indicated, diagnostic procedures, including ophthalmoscopy. Therefore, the perspective might be a shift from a “photo-assisted ophthalmoscopy” to an “ophthalmoscopy-assisted UWF fundus photography” approach.

A possible limitation of our study may be the use of this system without a deep learning tool, which may offer a more standardized classification.

In conclusion, we reported a substantial agreement in the classification of major retinal diseases using UWF imaging. This assessment highlights the interchangeability of UWF imaging and ophthalmoscopy by validating the technique of UWF imaging analysis in diagnosing major retinal diseases in daily clinical practice and should encourage the use of UWF imaging for fundus examination both in clinical and telehealth contexts. UWF imaging may improve the quality of clinical evaluation, allowing to compare images of the same eye acquired in different moments thus helping to monitor chronic diseases, such as diabetic retinopathy, and easily detect recurrences (i.e. choroidal and retinal tumors), also by using some tools intrinsic to the system. Moreover, it allows an accurate assessment of specific retinal areas, such as the macula or very peripheral sectors, without increasing the discomfort of the patient, as may happen during a prolonged ophthalmoscopic evaluation. The validation of UWF imaging for retinal evaluation also widens the possible uses of teleophthalmology: it allows to examine images at a different location than the clinic, possibly applying a deep learning algorithm, to reduce the time of evaluation and to perform screening for major retinal diseases thus rapidly directing the patient to a specific clinical care pathway if needed.