Cataract grading schemes have evolved since their implementation over 30 years ago. West and Taylor called for a standardized way to characterize and grade cataracts in 1986, and since then, the Oxford Clinical Grading System and early iterations of the Lens Opacities Classification System (LOCS) have contributed to the formation of the LOCS III system that is currently used today [1,2▪▪]. Grading systems are used both clinically and for the purposes of research, and the use of a grading schema plays an important role in both settings. Grading of cataracts may aid in the communication between physician and patient, as cataract grades can provide patients with understandable information about the diagnosis, progression, and importance of treatment for their cataracts. Additionally, grading a cataract allows the ophthalmologist to follow the cataract over a period of time to assess progression and to share data with other physicians in a clear and standardized way. A grading scheme also standardizes research findings, providing a uniform way to communicate findings and compare whereas a certain grade of cataract does not result in the same degree of visual impairment for each patient, it does provide insight into how severe vision impairment may be. As grading systems for cataracts have evolved, so has technology that allows clinicians to better assess, grade, and monitor cataract formation and progression. Grading systems combined with advancements in technology provide patients with increased information and improvements in their care.
HISTORICAL GRADING SYSTEMS
To understand cataract grading systems at present, it is helpful to reflect on the development of grading systems of the past. In the mid-1980s, a combination of resolution target projection ophthalmoscopy (acuity scope), the Oxford Clinical Grading system, and photographs were recommended to be used to assess cataracts . Although resolution target projection ophthalmoscopy is no longer used and photographic technologies have improved greatly, grading systems have remained a standard clinical tool. The Oxford Clinical Grading system of 1983 was created to be a comprehensive way to score a variety of cataract features and included the following elements: anterior clear zone thickness, anterior subcapsular opacity, posterior subcapsular opacity, cortical spoke opacity, waterclefts, vacuoles, retro-dots, focal dots, nuclear brunescence and white nuclear scatter . The features listed were given a grade on a scale from 0 to 5. There existed complexity behind each numerical grade; for example, ‘vacuoles [were] graded according to their frequency within an 8 mm diameter area by comparison with standard diagrams showing frequency on a logarithmic scale’ . Results of using the Oxford Clinical Grading system were mixed when tested among four physicians, as there was varying degree of agreeability among the various elements tested. Inter-observer mean weighted kappa values ranged from 0.87 (excellent) to 0.36 (poor), prompting authors of the 1987 article to caution users that there may be variability in the system .
Perhaps in contrast to the Oxford Clinical Grading System, several groups of researchers around this time set out to create a simplified grading scheme to increase efficiency in eye surveys and epidemiological studies of cataract patients, especially in rural communities. Mehra and Minassian  had four ophthalmologists and trained ophthalmic assistants grade lens opacity on a scale from 0 to 5 based on the amount of red reflex visualized by direct ophthalmoscope. Kappa values for this study ranged from 1.0 to 0.60, and there was good correlation between visual impairment and the grades given for cataracts, most likely owing to the simplicity of the grading system . Another study at the time also set out to create a model to take standardized slit-lamp photographs and to use a simplified 0–4 grading scheme for both nuclear and cortical cataracts . The resulting kappa values of this study also proved high reproducibility with this model . Lastly, the Beaver Dam Eye Study, referred to by most as the Wisconsin cataract grading system, grades nuclear sclerosis on a scale from 0 to 5 and cortical cataracts in a sum of nine separate lens segments . The Wisconsin cataract grading system is often used at present when comparing clinically used grading systems with new automated techniques in cataract grading . These studies provided an example that although much less complex than the Oxford Clinical Grading system, a simple grading scheme could correlate well with the patient's experience of their cataract. But while visual experience is a patient's primary concern, what did clinicians need out of their grading scheme in order to enhance patient care and treatment?
GRADING SCHEMES FROM PAST TO PRESENT
The LOCS was created with the goal to provide a reliable and reproducible grading scheme that was simpler for clinicians to use while also accounting for important cataract characteristics . Introduced in 1988, the LOCS I set out to classify cataracts using a slit lamp examination or with slit lamp and retroilluminated photographs . The LOCS I system evaluated the following: opacification in the cortical and posterior subcapsular zones and intensity of opalescence in the nucleus. Nuclear color and opalescence were evaluated separately, as it was found that color had less to do with cataract severity than previously thought . Of note, LOCS I marked a shift in focus from grading cataract severity based on visual acuity and instead basing it on cataract morphology, as unaggregated changes in the cortex had little role in visual acuity changes . Instead, clustered aggregation provided clinicians a way to tell a patient that they have an early cataract, independent of visual symptoms.
The LOCS II was developed the following year, improving upon LOCS I by implementing ways to differentiate degrees of cortical, subcapsular, and nuclear opacification in addition to adding color photographs to be used as standards for comparison . Visual acuity was again left out of the grading scheme, as there are other factors that contribute to changes in vision that could be unrelated to lens abnormality. LOCS II demonstrated good intra-observer and inter-observer agreements at the slit lamp and in photographic readings . The slit lamp exam proved to be slightly more sensitive at detecting opacification than photographs, likely owing to limitations in photographic technologies at the time .
The LOCS III was created in 1993 to improve upon limitations in LOCS II. LOCS III expanded the scales used in LOCS II to better capture the early stages of cataract formation in the grading scheme [2▪▪]. It examines nuclear opalescence (NO) and nuclear color (NC) on a scale from 1 to 6, cortical cataracts (C) on a scale from 1 to 5, and posterior subcapsular cataracts on a scale from 1 to 5 (Fig. 1) [2▪▪]. Although a finer grading scheme lowers concordance among measurements, the resulting increase in sensitivity was chosen as a worthy reason for the change. Of note, the original 1993 article highlighting improvements of LOCS III over LOCS II compared photographic evaluation of cataracts as opposed to in-vivo evaluation based on the ability to standardize photographic protocols and to reduce user variation in the slit lamp examination [2▪▪]. This comparison marked an important change in welcoming the idea that photographic assessment could be superior to the slit lamp exam, perhaps creating a gateway for digital analysis to provide a more objective measure of cataract than the subjective clinician assessment. At present, the LOCS III system is still used in varying degrees in clinical practice, however, its clinical impact on decision-making and timing of surgery is questionable.
CORRELATION BETWEEN GRADE AND VISUAL SYMPTOMS FOR THE PATIENT
Before discussing cataract grading systems further, an important question to ask is how it relates to the symptoms experienced by the patient. If cataract surgery is to be performed regardless of cataract severity and instead based on patient experience of their cataract, what benefit does a comprehensive grading scheme provide? From an epidemiological standpoint, it is important to characterize cataract type and severity to better understand population health and treatment need, but how do cataract grades perform on a patient-to-patient basis? In a 2012 study examining LOCS III grading and visual functioning, it was noted that vision-specific functioning decreased in a statistically significant way for different cataract types at different grades using the LOCS III [9▪]. Overall, the trend makes sense in that as grade increased, vision-specific functioning decreased. The nuance, however, lies in the fact that each cataract type produced symptoms at different grades. Nuclear cataracts affected functioning at grades 4 for opalescence and 5 for color, cortical cataracts affected functioning at grade 3, and posterior subcapsular cataracts affected functioning at all grades [9▪]. Although it may make sense to a clinician that a posterior subcapsular cataract produces visual symptoms in a more significant way than does a nuclear cataract, from a patient perspective it may be confusing to find that a cataract grade of P1 is much more visually bothersome than a cataract grade of NC3 if the patient is unfamiliar with the nuances of cataract disease. The study raises an important consideration of whether or not vision-specific functioning should be incorporated back into the assessment of cataracts if the grading scheme is not only meant to incorporate physician findings but also patient perspective and understanding.
CORRELATION BETWEEN GRADE, SURGICAL APPROACH, AND OUTCOME
Not only is it important to consider the patient's experience of their cataracts and how it relates to cataract grade, it is also important to consider how the surgical approach and outcome are determined by cataract grading. Will surgical approach be different if a cataract is given a grade of NO3 and NC3 versus NO4 and NC4? A study by Davidson and Chylack investigating the use of LOCS III and phacoemulsification performance in 2364 cases found that nuclear cataract phacoemulsification time correlated well with LOCS III but that cortical and posterior subcapsular phacoemulsification had little relation to LOCS III classes [10▪▪]. Of note, an exponential increase in phacoemulsification energy was used intraoperatively as nuclear grades increased, proving LOCS III to be a useful tool in creating an operative plan for nuclear cataract procedures [10▪▪]. Cataract scores can play a useful role in determining case complexity during cataract surgery and postoperative outcomes. This preoperative knowledge can better allow clinicians to allocate certain cases to surgeons in training and to provide realistic risks and potential outcomes to patients.
NEW WAYS TO ASSESS AND GRADE CATARACTS
Since the development of the LOCS III system, there have been a variety of ways presented to better evaluate and grade cataracts. Shortly after the LOCS III system was introduced, Hall et al. set out to provide a new way to evaluate nuclear cataracts using a laser slit lamp in an attempt to standardize an objective way to visualize cataract complexities. The team created a device using a laser light slit, viewing arm with a beam splitter, and a charge coupled device camera to illuminate the anterior segment with laser light . The images obtained with the laser were analyzed with a computer program that calculated mean pixel intensity of the area of interest in the nucleus as compared with a darker background portion of the anterior segment in the same image to measure the amount of light backscattered from the nucleus . This method proved to correlate well with the LOCS III grading scheme, demonstrating a linear relationship between nuclear opalescence LOCS III scores and the pixel intensities imaged by the laser slit lamp and analyzed with computer software . This highlighted the use of a computer program to assist with cataract grading.
Following the work of Hall, Babizhayev et al.  set out to use computer generated analysis of lens imaging to measure cataract severity in combination with a glare disability test. An automated computer program was used to measure the intensity of individual image pixels, and using serial 2D images, 3D topography images of the lens were created to provide a better understanding of the characteristics of the lens . The computer's analysis of the lens was compared against a clinical slit lamp examination using the same grading scheme laid out by Taylor and West. This analysis technique demonstrated correlation between the existing LOCS III scoring system and a new way to assess cataracts .
Additional teams of researchers further refined automatic systems for classifying cataracts using slit lamp photography. Fan et al.  created a method to grade nuclear sclerotic cataracts based on landmarks in the visual axis from slit lamp photographs. Another group from Johns Hopkins also used photographs from the visual axis and analyzed a variety of features of the lens . In these methods, a neural network was employed to analyze the images and produce a grade of nuclear cataract severity based on the lens landmarks noted in the visual axis . At the time, these models focused only on the visual axis, whereas a clinical exam evaluated the entirety of the lens.
Li et al.  expanded on previous groups’ work to create an automatic diagnosis system for nuclear cataracts by training the system to locate the entire lens and to extract features such as intensity, color, and entropy inside the lens and nucleus. They based the lens structure detection ability of their program on their prior work with the Active Shape Model method that they described in previous articles, and they built upon it to better identify the lens nucleus in addition to the lens itself [16,17]. The lens and feature detection system was tested on over 5800 images with a 96.8% lens location detection rate and a 95% lens structure detection rate . When compared against clinical diagnoses of the same images, the automatic diagnosis system only varied by greater than 1 clinical grade in only 3.37% of images reviewed, proving its agreement with clinical grading schemes and clinical diagnostic methods . Xu et al.  set out to create an automatic grading system that refined the work of Li by building on the success of the lens structure detection and identifying the most effective elements from the image. This bag-of-features (BOF) method allows for location-independent representation of features that can be assessed for qualities such as intensity and scale. Combining the BOF method with regression model training, the system achieved a 69.0% exact agreement ratio when compared with clinical grading and a 98.9% decimal grading error less than 1.0 . These results demonstrated that clinical diagnosis could be facilitated with the use of automatic diagnosis systems.
Improving upon these automatic diagnosis systems, Srivastava et al.  developed what they call the Automatic Cataract Screening from Image Analysis-Nuclear Cataract, Version 0.10, ACASIA-NC_v0.10. The system builds upon the work done by Li et al. in that it uses image gradients on edges produced by landmarks in the lens. At higher grades of nuclear cataract, these landmarks tend to be less visible as the landmarks become less distinct. The ACASIA-NC_v0.10 instead focuses on the gradient between areas of the lens as a means to measure the difference in grades of cataract and to overcome the loss of visible landmarks in severe cataracts . Although the ACASIA-NC_v0.10 did not prove to be useful in grading very low or very high grade nuclear cataracts, it improved upon prior methods that used color and intensity and demonstrated that gradient information is useful in automatic cataract grading programs .
BEYOND SLIT LAMP PHOTOGRAPHY: NEW TECHNIQUES IN CATARACT ASSESSMENT AND GRADING
Although slit lamp examination and slit lamp photographs have been the mainstay of cataract grading systems, different ways to assess for cataracts have been proposed. These methods may vary in the exactness of which they can assess cataracts as compared with standard slit lamp photography analysis, but they may provide alternatives to cataract grading when resources or time are limited. Using imaging modalities in addition to slit lamp photographs, Xiong et al. [20▪▪] suggest that retinal images can be used to screen for cataracts. They employed the scale used by Wang et al.  to grade blurriness in retinal images. Previous work on the topic was performed with Fourier analysis by Abdul-Rahman et al. . Their automated method of quantifying optical degradation in retinal images proved to be useful in detecting the presence of cataract and correlated well with LOCS III . However, automatic grading of the cataract was not performed. Additionally, vitreous opacity could not be evaluated separately from lens opacities, potentially influencing grading . With these limitations in mind, Xiong et al. developed a successful method to detect vitreous opacity and separate it from the retinal structure detection so as to improve cataract grading accuracy. This system graded cataracts with an 81.1% accuracy and kappa value of 0.7435 when compared with clinical grading [20▪▪]. As technologies in telemedicine and artificial intelligence advance, using retinal images to grade cataract may prove an efficient and useful tool to utilize fundus photos that may already be taken for diabetic retinopathy or age-related macular degeneration monitoring.
In addition to fundus photography being used as a novel method to grade cataracts, optical coherence tomography (OCT) has also been suggested. Anterior segment OCT has been used to assess a variety of features such as corneal thickness, but Wong et al. thought to compare the anterior segment OCT nucleus density measurement with LOCS III grading of nuclear opalescence and nuclear color. They found that there was a significant correlation between the two measurements and that nuclear opalescence scores had a slightly higher association with anterior segment OCT than did nuclear color . Their use of anterior segment OCT to grade lens density proved to be an objective, reliable, and fast assessment using a frequently used clinical tool that requires less training than mastering the LOCS III.
THE ROLE OF DEEP LEARNING IN CATARACT GRADING
A discussion of cataract grading would be incomplete without touching on recent advancements in the use of deep learning with machine learning and artificial intelligence to assess lens opacity. Although the existing automatic methods to grade cataracts discussed previously rely on predefined landmarks and features to recognize structures and grade cataracts, systems can be trained to learn grading features, filter them, and feed them into a neural network to analyze them further. Gao et al. [24▪▪] used this approach with a 5378 slit lamp image data set and obtained a 70.7% exact integral agreement ratio and a 99.0% decimal grading error less than 1.0 when compared with clinical integral grading. They demonstrated that deep learning can be an effective and powerful tool to grade cataracts.
The shift towards using computer-aided analysis of cataract imaging to better standardize cataract grading is an exciting and rapidly developing field. The role for automatic grading of cataracts is growing as new technologies are developed to image the eye and to better view structures that were previously hard to capture on imaging. With the grading systems of past in mind, researchers have been able to build on the classical ways that cataracts have been assessed through slit lamp examination and use the features of the lens that are typically assessed as a platform for automated systems to grade cataracts in an objective and standardized way.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
REFERENCES AND RECOMMENDED READING
Papers of particular interest, published within the annual period of review, have been highlighted as:
- ▪ of special interest
- ▪▪ of outstanding interest
1. West SK, Taylor HR. The detection and grading of cataract: an epidemiologic perspective. Surv Ophthalmol 1986; 31:175–184.
2▪▪. Chylack LT, Wolfe JK, Singer DM, et al. The lens opacities classification system III
. Arch Ophthalmol 1993; 111:831–836.
The LOCS III system defined in this publication is a fundamental clinical diagnostic tool and is used frequently as the basis of comparison for developing technologies that diagnose and grade cataracts.
3. Sparrow JM, Bron AJ, Brown NA, et al. The Oxford Clinical Cataract Classification and Grading System. Int Ophthalmol 1986; 11:151–157.
4. Mehra V, Minassian DC. A rapid method of grading cataract in epidemiological studies and eye surveys. Br J Ophthalmol 1988; 72:801–803.
5. Taylor HR, West SK. The clinical grading of lens opacities. Aust N Z J Ophthalmol 1989; 17:81–86.
6. Klein BEK, Klein R, Linton KLP, et al. Assessment of cataracts from photographs in the Beaver Dam Eye Study. Ophthalmology 1990; 97:1428–1433.
7. Chylack LT, Leske MC, Sperduto R, et al. Lens opacities classification system. Arch Ophthalmol 1988; 106:330–334.
8. Chylack LT, Khu P, Kashiwagi T, et al. Lens opacities classification system II (LOCS II). Arch Ophthalmol 1989.
9▪. Chew M, Chiang PPC, Zheng Y, et al. The impact of cataract, cataract types, and cataract grades on vision-specific functioning using rasch analysis. Am J Ophthalmol 2012; 154:29.e2–38.e2.
In this study, the authors show that vision-specific functioning decreased in a statistically significant way for different cataract types at different grades using the LOCS III, demonstrating that each cataract type produced symptoms at different grades.
10▪▪. Davison JA, Chylack LT. Clinical application of the Lens Opacities Classification System III
in the performance of phacoemulsification. J Cataract Refract Surg 2003; 29:138–145.
This study demonstrated a positive exponential correlation between phacoemulsification energy requirements during surgery and nuclear cataract grade using the LOCS III but did not show a correlation for posterior subcapsular or cortical cataracts.
11. Hall NF, Lempert P, Shier RP, et al. Grading nuclear cataract: Reproducibility and validity of a new method. Br J Ophthalmol 1999; 83:1159–1163.
12. Babizhayev MA, Deyev AI, Yermakova VN, et al. Image analysis and glare sensitivity in human age-related cataracts. Clin Exp Optom 2003; 86:157–172.
13. Fan S, Dyer CR, Hubbard L, Klein B. Ellis RE, Peters TM. An automatic system for classification of nuclear sclerosis from slit-lamp photographs. Medical image computing and computer-assisted intervention - MICCAI 2003. Berlin, Heidelberg: Springer Berlin Heidelberg; 2003. 592–601.
14. Duncan DD, Shukla OB, West SK, Schein OD. New objective classification system for nuclear opacification. J Opt Soc Am A Opt Image Sci Vis 1997; 14:1197–1204.
15. Li H, Lim JH, Liu J, et al. An automatic diagnosis system of nuclear cataract using slit-lamp images. Conf Proc IEEE Eng Med Biol Soc 2009; 2009:3693–3696.
16. Li H, Lim JH, Liu J, Wong TY. Towards automatic grading of nuclear cataract. In: Annual International Conference of the IEEE Engineering in Medicine and Biology – Proceedings. 2007.
17. Li H, Lim JH, Liu J, et al. Image based grading of nuclear cataract by SVM regression. Proc SPIE 6915
, Medical Imaging 2008: Computer-Aided Diagnosis, 691536 (17 March 2008).
18. Xu Y, Gao X, Lin S. Automatic grading of nuclear cataracts from slit-lamp lens images using group sparsity regression. In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 2013.
19. Srivastava R, Gao X, Yin F, et al. Automatic nuclear cataract grading
using image gradients. J Med Imaging (Bellingham) 2014; 1:014502.
20▪▪. Xiong L, Li H, Xu L. An approach to evaluate blurriness in retinal images with vitreous opacity for cataract diagnosis. J Healthc Eng 2017; 2017:5645498.
This study goes beyond slit lamp photography and proposes a method of detecting and grading cataracts with high success using retinal images that may already need to be obtained for other clinical purposes, perhaps improving clinical efficiency.
21. Yang Chuan-wu, Xu Liang, Wang Shuang, et al. The evaluation of screening for cataract needed surgery with digital nonmydriatic fundus camera. Ophthalmol China 2010; 19:46–49.
22. Abdul-Rahman AM, Molteno T, Molteno ACB. Fourier analysis of digital retinal images in estimation of cataract severity. Clin Exp Ophthalmol 2008; 36:637–645.
23. Wong AL, Leung CK-S, Weinreb RN, et al. Quantitative assessment of lens opacities with anterior segment optical coherence tomography
. Br J Ophthalmol [Internet] 2009; 93:61–65.
24▪▪. Gao SL X, TYW. Automatic feature learning to grade nuclear cataracts based on deep learning
. IEEE Trans Biomed Eng 2015; 62:2693–2701.
This study highlights the use of deep learning as a powerful and highly reproducible tool to grade cataracts.