Automatic image analysis of episcleral hemangioma applied to the prognosis prediction of trabeculotomy in Sturge–Weber syndrome☆
Introduction
Sturge–Weber syndrome (SWS) is a rare neurocutaneous disorder with an incidence ranging from 1 in 50,000 to 1 in 20,000 [1]. It is a sporadic, congenital, neurocutaneous disorder with angiomas that is characterized by leptomeningeal angiomatosis, glaucoma, and cutaneous capillary malformations in the distributions of the trigeminal nerve [2]. Nowadays, trabeculotomy has some chance of curing young SWS with glaucoma [3], [4]. To improve the recovery rate, there is an urgent need to find indexes that characterize or predict the success of the surgery. Compared to other organs, the blood vessels of the eyes are visible and can indicate the health condition of the body organs [5]. SWS is a disease characterized by vascular malformations. In previous studies, episcleral hemangiomas were found in all patients with secondary glaucoma [6]. Therefore, analyzing the blood vessels in the eyes seems to provide a window for us to evaluate glaucoma. However, the researchers did not find an association between scleral hemangioma and glaucoma outcome.
Inspired by this, we observed the vascular distribution patterns in dozens of SWS patients, showing that the prognosis of glaucoma after trabeculotomy in SWS patients was closely related to the distribution pattern of scleral hemangioma in the eyes [7]. We had established a scoring system to distinguish the degree of vascular malformation. The more severe the vascular malformation of the scleral surface, the higher the score. The artificial scores were used to divide episcleral blood vessels into two categories: namely MEVAN (multiple episcleral vascular abnormal network) and SEVAN (simple episcleral vascular abnormal network) [7]. In our study, 24 of the 50 eyes (46 patients) were assigned to MEVAN, and the remaining 26 were assigned to SEVAN. In the subsequent two years of follow-up, the trabeculotomy success rate of the SEVAN group was 89.7%, and the MEVAN group was 36%. The cumulative success rates were significantly different between the 2 groups (). Therefore, we concluded that the vascular distribution pattern could be an indicator of prognosis.
Our previous study was mainly based on the observation results of ophthalmologists from Shanghai Ninth People’s Hospital. Since SWS is a rare disease, the quantitative lacking-manual scoring system might be affected by doctors’ experience and other subjective factors. Hence, in the current work, we attempted to apply an image analysis algorithm to extract image features representing the distribution characteristics of episcleral vessels, and afterward used an unsupervised clustering method to classify the SWS patients into two categories namely the successful operation case and the failed operation case based on the extracted image features. The survival analysis was conducted to explore the relationship between clustering results and surgical prognosis. Automatic diagnosis by the designed algorithm is of great significance to the process of auxiliary medical diagnosis, and is not easily affected by the subjective experience of ophthalmologists.
Section snippets
Experimental data
The experimental data in this study were obtained from SWS patients under 4 years old who underwent trabeculotomy in Shanghai Ninth People’s Hospital Affiliated to Shanghai Jiao Tong University School of Medicine from February 2016 to July 2017. The study was approved by the Review Committee of the Ninth People’s Hospital Affiliated to Shanghai Jiao Tong University School of Medicine. Diagnosis, surgery, follow-up, and necessary treatment were provided by the ophthalmologist (Dr. Guo Wenyi). We
Segmentation results of episcleral vessels
Fig. 2 demonstrates the original image of the episcleral vessels (first line) and the binary image of the episcleral vessels (second line). As shown in Fig. 2, most blood vessels in images can be segmented by the selected threshold-based segmentation algorithm. In general, the overall structure of the segmented vessels is well maintained. At the same time, there is less discrete noise in segmented images, and this will facilitate the subsequent statistical image analysis such as image feature
Discussion
In previous image analysis of ophthalmic vascular disease, diabetic retinopathy, hypertension, and glaucoma were mainstream research objects [23]. The goal of these previous researches was to automatically and objectively distinguish different stages of the disease development using image analysis technique. For the standard datasets of these ophthalmic vascular diseases, by extracting image features specific to disease development characteristics, the classification accuracy of the
Conclusions
In conclusion, two image features viz. entropy and density extracted from episcleral blood vessel image can be used as indicators to automatically and objectively evaluate the severity of episcleral hemangioma, the severity of glaucoma and to predict trabeculotomy outcomes of young SWS patients. Guided by these two image features, ophthalmologists may develop the effective surgical strategy to raise the success rate of SWS treatment.
CRediT authorship contribution statement
Lejing Zhang: Methodology, Validation, Writing – original draft, Writing – review & editing. Yue Wu: Funding acquisition, Methodology, Validation, Writing – original draft, Writing – review & editing. Menghan Hu: Funding acquisition, Methodology, Validation, Writing – original draft, Writing – review & editing. Wenyi Guo: Writing – review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work is sponsored by the Interdisciplinary Program of Shanghai Jiao Tong University, China (YG2019QNA18), the “Chenguang Program” of Shanghai Education Development Foundation, China and Shanghai Municipal Education Commission, China (No. 19CG27), the Science and Technology Commission of Shanghai Municipality, China (No. 19511120100), the GHfund B (No. 20210702), and the Fundamental Research Funds for the Central Universities, China .
References (31)
- et al.
Early trabeculotomy ab externo in treatment of sturge-Weber syndrome
Am. J. Ophthalmol.
(2017) - et al.
Medical device design in context: A model of user–device interaction and consequences
Displays
(2012) - et al.
Large-scale elemental image array generation in integral imaging based on scale invariant feature transform and discrete viewpoint acquisition
Displays
(2021) - et al.
Image inpainting based on deep learning: A review
Displays
(2021) - et al.
New vascular classification of port-wine stains: improving prediction of sturge Weber risk
Br. J. Dermatol.
(2014) - et al.
Screening for sturge-Weber syndrome: A state-of-the-art review
Pediatr. Dermatol.
(2017) - et al.
Long-term outcomes of primary trabeculotomy for congenital glaucoma at children’s medical center
Invest. Ophthalmol. Vis. Sci.
(2010) - E. Deepika, S. Maheswari, Earlier glaucoma detection using blood vessel segmentation and classification. In: 2nd...
Dual origin of glaucoma in encephalotrigeminal haemangiomatosis
Trans. Ophthalmol. Soc. United Kingd.
(1973)- et al.
Episcleral hemangioma distribution patterns could be an indicator of trabeculotomy prognosis in young SWS patients
Acta. Ophthalmol.
(2020)
Automated techniques for blood vessels segmentation through fundus retinal images: A review
Micros. Res. Tech.
Retinal blood vessel segmentation using line operators and support vector classification
IEEE Trans. Med. Imaging
Blood vessel segmentation from color retinal images using unsupervised texture classification
IEEE Int. Conf. Image Process.
Cited by (0)
- ☆
This paper was recommended for publication by Zhi-Cheng Li.