Skip to content

Classification and grading

TL;DR — Cataract is graded by anatomic type — nuclear opalescence (NO), nuclear colour (NC), cortical (C) and posterior subcapsular (P) — and by severity on a photographic reference scale. LOCS III remains the reference standard: it replaced LOCS II's integer scale with decimalised standards and narrowed 95% tolerance limits from 2.0 units to 0.7 (NO), 0.7 (NC), 0.5 (C) and 1.0 (P) (Chylack 1993, PMID 8512486). Its weakness is intrinsic: it is a subjective slit-lamp comparison, rarely used clinically, and it correlates only moderately with what the patient can see (Mackenbrock 2024, PMID 38242135). Objective alternatives — Scheimpflug densitometry, AS-OCT, autofluorescence, deep learning — reproduce LOCS III grades well (agreement ρ ≈ 0.47 for Pentacam vs LOCS III; AUC 0.92–1.00 for CNN classifiers) but no objective scale has replaced it as the field's common currency (Mirzaie 2022, PMID 36160089; Tang 2025, PMID 40995559). For field epidemiology the WHO simplified three-level system exists precisely because LOCS III is too demanding for survey conditions (Thylefors 2002, PMID 11821974). The unresolved measurement problem is that no grading system is calibrated against the outcomes that trigger surgery — function, glare and contrast.

Anatomic phenotypes

Type Location Typical optical effect Characteristic associations
Nuclear Central embryonic/fetal/adult nucleus; progressive sclerosis and brunescence Reduced contrast, myopic shift, reduced colour discrimination Age; the lens grows throughout life so the core is exposed longest, and a barrier to glutathione transport forms around the nucleus in the fourth decade (Michael 2011, PMID 21402586; Truscott 2005, PMID 15862178)
Cortical Wedge/spoke-shaped opacities in the equatorial cortex extending inward Glare and scatter, often disproportionate to Snellen acuity Cumulative UV-B exposure (Taylor 1988, PMID 3185661)
Posterior subcapsular (PSC) Plaque immediately anterior to the posterior capsule, on the visual axis Marked loss of near vision and of vision in bright light (pupil constriction) Corticosteroids, diabetes, uveitis, radiation, high myopia (Prokofyeva 2013, PMID 22715900; Świerczyńska 2025, PMID 39813957)
Mixed Two or more of the above Combination 22% of incident cataract in the Blue Mountains cohort had >1 type; 1.3% had all three (Kanthan 2008, PMID 17900695)

Because the three phenotypes have partly different risk factors and different symptom signatures, an aggregate "cataract" prevalence figure mixes at least three diseases. Cortical opacity in particular can degrade function through scatter while leaving high-contrast acuity nearly intact.

LOCS III and its predecessors

The Lens Opacities Classification System was built to make clinical grading reproducible enough for longitudinal research. LOCS II was shown to be reliable but its performance depended on the lesion: severity of cortical and nuclear opacity did not affect slit-lamp reproducibility, whereas clinical grading of PSC became more reliable as PSC severity increased, and more advanced coexisting opacities degraded agreement for nuclear (but not cortical or PSC) diagnosis in 3,646 eyes of the Italian-American Natural History Study (Maraini 1991, PMID 2071351).

LOCS III addressed four named limitations of LOCS II — unequal intervals between standards, a single standard for colour grading, integer-only grading, and wide tolerance limits. It provides six slit-lamp images for NC and NO, five retroillumination images for C and five for P, on a decimal scale with regularly spaced intervals (PMID 8512486). Reported reliability from the same group: kappa 0.85–1.00 for LOCS II and intraclass correlation 0.67–0.94 for LOCS III, against objective comparators with much higher internal reliability — fast spectral scanning colorimetry r₁ = 0.96–0.98, nuclear mean density r₁ = 0.97, percent area opacity r₁ = 0.92–0.96 (Chylack 1993, PMID 8302524).

LOCS III is field-usable. In 150 subjects aged 33–55 attending an Indian refraction clinic, interrater reliability was high and a change of ≥0.5 units in colour, cortical, nuclear or PSC grade was seen in at least one eye of 54% of subjects over one year (Srinivasan 1997, PMID 9486033).

System Structure Design intent Source
LOCS II Integer grades, one colour standard First widely adopted reproducible clinical system PMID 2071351
LOCS III 6 NC/NO + 5 C + 5 P standards, decimal scale; 95% tolerance limits 0.7/0.7/0.5/1.0 Narrower tolerance, finer resolution, longitudinal use PMID 8512486
Oxford Clinical Cataract Classification, decimalised 10 graded features, decimal scale Increase precision and responsiveness; repeatability "good to excellent" in 217 paired observations Sparrow 2000, PMID 10652171
WHO simplified system Three severity levels for nuclear, cortical, PSC; 3 standard photos for nuclear Field use in prevalence surveys by minimally trained graders; interobserver agreement "very good to fair" at four sites Thylefors 2002, PMID 11821974
Wisconsin system Photographic grading of lens photographs Used for incidence estimation in population cohorts PMID 17900695

Reviews of the lineage note the same tension: LOCS III is the research gold standard, is limited by its subjective nature, and is rarely used in ordinary clinics (Gali 2019, PMID 30489359; Mackenbrock 2024, PMID 38242135).

Objective and image-based grading

Modality Metric Performance reported Source
Scheimpflug (Pentacam PCGS/PNS) Lens densitometry, Pentacam Nucleus Staging Correlation with LOCS III grade ρ = 0.47 (P < 0.001) in 300 pure senile cataract eyes; Bland–Altman showed only moderate alignment Mirzaie 2022, PMID 36160089
Optical line-spread function Width of blurred line image (WSCAT) Distinguished eyes with 6/4.5–6/6 acuity from eyes ≤6/9 (95% CI), giving an optical rather than morphological measure Karbassi 1993, PMID 8302532
Objective scatter/dysfunction indices (OSI, DLI, PNS) Screening cut-offs for contrast recovery after multifocal IOL Cut-offs age ≤62, OSI ≤1.25, DLI ≥7.67, PNS ≤ threshold; contrast-sensitivity defocus curve correlated with objective metrics better than acuity did Fernández 2023, PMID 36871115
Ultra-wide-field fundus autofluorescence Mean grey-value difference Rose monotonically across NC/NO grades 1→6 (53.3 ± 11.4 to 121.1 ± 12.0; P < 0.001) in 60 eyes Eom 2023, PMID 36826601
AS-OCT + CNN (AFSNet) Nuclear severity class Outperformed strong baselines on a clinical AS-OCT dataset Zhang 2022, PMID 35245700
Slit-lamp video + machine learning WHO nuclear grade 206,574 frames from 1,812 eyes; AUC 0.967 (NUC 0), 0.928 (1), 0.923 (2), 0.949 (3) Shimizu 2023, PMID 38086904
Retroillumination image analysis Cortical opacity area ROI detection success 98.2% of 611 images; exact grade agreement 85.6% of 466 images; mean opacity-area error 3.15% vs human grader Li 2008, PMID 19163566
Hybrid CNN on LOCS III classes NO/NC/C/P classification Accuracy 90.88–100%, AUC 96.68–100% Tang 2025, PMID 40995559

The pattern across these is consistent: objective systems have higher internal reliability than the standard they are validated against, which caps measurable agreement and makes "correlation with LOCS III" a weak criterion. Reviews of AI in cataract identify the same limitations — small non-diverse training sets and little external validation (Goh 2020, PMID 32349116).

Grading, surgical difficulty and outcome

Grade is not only descriptive. Cataract grade predicts operative work and intraoperative behaviour: in an intra-individual comparison of femtosecond-laser-assisted versus conventional phacoemulsification in the same patients matched for LOCS III grade, total pupil variation differed significantly for grade ≤3 (0.08 ± 0.22 mm², P = 0.034) but not for grade >3 (0.01 ± 0.23 mm², P = 0.849), and cumulative dissipated energy differed significantly between techniques (P < 0.001) (Salgado 2023, PMID 37551374). Objective densitometry correlates with phaco energy and surgical time, which is why device-based grading is drifting from a research tool toward a planning tool (PMID 38242135).

Where grading fails

  • It does not measure disability. No grading system is calibrated against function; visual acuity itself is a poor gauge of cataract disability, and glare, contrast and light-scatter measures capture things acuity misses (See 2019, PMID 30489358; PMID 8302532).
  • Mixed cataract has no agreed attribution rule. With 22% of incident cataract being multi-type (PMID 17900695), assigning a person to "nuclear" or "cortical" for risk-factor analysis is a modelling decision that varies between studies.
  • Threshold choice moves prevalence. Survey estimates shift roughly two-fold with the acuity threshold used to define cataract blindness (Song 2018, PMID 29977532); grading thresholds have the same leverage.
  • No cross-system calibration exists. LOCS III, Oxford, WHO simplified, Wisconsin and device scales have never been mapped onto one another with published conversion functions and uncertainty, so pooled prevalence and pooled progression estimates rest on an unstated assumption of equivalence (PMID 11821974; PMID 10652171; PMID 30489359).

Open questions

  • Can grading systems be formally cross-calibrated? LOCS III, the decimalised Oxford system, the WHO simplified system and Scheimpflug/AS-OCT metrics are all in current use (PMID 8512486; PMID 10652171; PMID 11821974; PMID 38242135), but no study has published bidirectional conversion functions with prediction intervals — so burden estimates pooled across systems carry an unquantified error.
  • Does an objective grade predict patient-important outcomes better than acuity? Objective scatter indices already outperform acuity for predicting contrast-sensitivity recovery after multifocal implantation in one retrospective series (PMID 36871115); no prospective study has tested a densitometry threshold against a PROM-based indication.
  • What is the external validity of deep-learning graders? Reported AUCs are 0.92–1.00 on internal datasets (PMID 38086904; PMID 40995559), yet reviews specifically flag the absence of robust external validation and diverse training data (PMID 32349116); no cross-population external validation with prespecified thresholds has been reported.
  • Is grader-based progression a usable trial endpoint? A ≥0.5 LOCS III unit change occurred in 54% of an Indian cohort within a year (PMID 9486033), but the minimum clinically important difference in LOCS III units — the change a patient would notice — has never been established.

References

  1. Chylack LT, Wolfe JK, Singer DM, et al. The Lens Opacities Classification System III. The Longitudinal Study of Cataract Study Group. Archives of ophthalmology (Chicago, Ill. : 1960). 1993;111:831-6. PMID 8512486
  2. Mackenbrock LHB, Labuz G, Baur ID, et al. Cataract Classification Systems: A Review. Klinische Monatsblatter fur Augenheilkunde. 2024;241:75-83. PMID 38242135
  3. Mirzaie M, Bahremani E, Taheri N, et al. Cataract Grading in Pure Senile Cataracts: Pentacam versus LOCS III. Journal of ophthalmic & vision research. 2022;17:337-343. PMID 36160089
  4. Tang G, Zhang J, Du Y, et al. Artificial intelligence in cataract grading system: a LOCS III-based hybrid model achieving high-precision classification. Frontiers in cell and developmental biology. 2025;13:1669696. PMID 40995559
  5. Thylefors B, Chylack LT, Konyama K, et al. A simplified cataract grading system. Ophthalmic epidemiology. 2002;9:83-95. PMID 11821974
  6. Michael R, Bron AJ. The ageing lens and cataract: a model of normal and pathological ageing. Philosophical transactions of the Royal Society of London. Series B, Biological sciences. 2011;366:1278-92. PMID 21402586
  7. Truscott RJ. Age-related nuclear cataract-oxidation is the key. Experimental eye research. 2005;80:709-25. PMID 15862178
  8. Taylor HR, West SK, Rosenthal FS, et al. Effect of ultraviolet radiation on cataract formation. The New England journal of medicine. 1988;319:1429-33. PMID 3185661
  9. Prokofyeva E, Wegener A, Zrenner E. Cataract prevalence and prevention in Europe: a literature review. Acta ophthalmologica. 2013;91:395-405. PMID 22715900
  10. Świerczyńska M, Tronina A, Smędowski A. Understanding cataract development in axial myopia: The contribution of oxidative stress and related pathways. Redox biology. 2025;80:103495. PMID 39813957
  11. Kanthan GL, Wang JJ, Rochtchina E, et al. Ten-year incidence of age-related cataract and cataract surgery in an older Australian population. The Blue Mountains Eye Study. Ophthalmology. 2008;115:808-814.e1. PMID 17900695
  12. Maraini G, Pasquini P, Sperduto RD, et al. The effect of cataract severity and morphology on the reliability of the Lens Opacities Classification System II (LOCS II). Investigative ophthalmology & visual science. 1991;32:2400-3. PMID 2071351
  13. Chylack LT, Wolfe JK, Friend J, et al. Quantitating cataract and nuclear brunescence, the Harvard and LOCS systems. Optometry and vision science : official publication of the American Academy of Optometry. 1993;70:886-95. PMID 8302524
  14. Srinivasan M, Rahmathullah R, Blair CR, et al. Cataract progression in India. The British journal of ophthalmology. 1997;81:896-900. PMID 9486033
  15. Sparrow NA, Frost NA, Pantelides EP, et al. Decimalization of The Oxford Clinical Cataract Classification and Grading System. Ophthalmic epidemiology. 2000;7:49-60. PMID 10652171
  16. Gali HE, Sella R, Afshari NA. Cataract grading systems: a review of past and present. Current opinion in ophthalmology. 2019;30:13-18. PMID 30489359
  17. Karbassi M, Magnante PC, Wolfe JK, et al. Objective line spread function measurements, Snellen acuity, and LOCS II classification in patients with cataract. Optometry and vision science : official publication of the American Academy of Optometry. 1993;70:956-62. PMID 8302532
  18. Fernández J, Burguera N, Rocha-de-Lossada C, et al. Objective cataract grading methods and expected contrast sensitivity reestablishment with multifocal intraocular lenses. International ophthalmology. 2023;43:2825-2832. PMID 36871115
  19. Eom Y, Suh YW, Kim SW, et al. New technology using crystalline lens autofluorescence for presbyopia and cataract grading. Graefe's archive for clinical and experimental ophthalmology = Albrecht von Graefes Archiv fur klinische und experimentelle Ophthalmologie. 2023;261:1923-1932. PMID 36826601
  20. Zhang X, Xiao Z, Higashita R, et al. Adaptive feature squeeze network for nuclear cataract classification in AS-OCT image. Journal of biomedical informatics. 2022;128:104037. PMID 35245700
  21. Shimizu E, Tanji M, Nakayama S, et al. AI-based diagnosis of nuclear cataract from slit-lamp videos. Scientific reports. 2023;13:22046. PMID 38086904
  22. Li H, Ko L, Lim JH, et al. Image based diagnosis of cortical cataract. Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference. 2008;2008:3904-7. PMID 19163566
  23. Goh JHL, Lim ZW, Fang X, et al. Artificial Intelligence for Cataract Detection and Management. Asia-Pacific journal of ophthalmology (Philadelphia, Pa.). 2020;9:88-95. PMID 32349116
  24. Salgado R, Torres P, Marinho A, et al. Cataract Grade and Pupil: Comparison Between Conventional Phacoemulsification and Low-Energy Femtosecond Laser Assisted Cataract Surgery. Clinical ophthalmology (Auckland, N.Z.). 2023;17:2193-2200. PMID 37551374
  25. See CW, Iftikhar M, Woreta FA. Preoperative evaluation for cataract surgery. Current opinion in ophthalmology. 2019;30:3-8. PMID 30489358
  26. Song P, Wang H, Theodoratou E, et al. The national and subnational prevalence of cataract and cataract blindness in China: a systematic review and meta-analysis. Journal of global health. 2018;8:010804. PMID 29977532