Kellgren-Lawrence Grading System: A Complete Guide
The KL grading system for knee OA — radiographic criteria, clinical interpretation, inter-observer variability, and AI-assisted grading.
What Is the Kellgren-Lawrence Grading System?
The Kellgren-Lawrence (KL) grading system, first described by John Kellgren and John Lawrence in 1957, remains the most widely used radiographic classification for osteoarthritis severity. Originally published in the Annals of the Rheumatic Diseases, it provides a standardised 5-point scale (Grade 0–4) based on anteroposterior (AP) weight-bearing knee radiographs.
Despite being nearly seven decades old, KL grading is still the de facto standard in clinical trials, epidemiological studies, and routine orthopaedic practice. Understanding its strengths and limitations is essential for any surgeon working with OA patients.
The Five Grades Explained
Grade 0 — Normal. No radiographic features of osteoarthritis. Joint space is preserved, no osteophytes, no sclerosis. This is the baseline against which all other grades are compared.
Grade 1 — Doubtful. Possible osteophytic lipping at the joint margin. Joint space is not definitively narrowed. This grade has the highest inter-observer variability — two radiologists looking at the same film may disagree on whether subtle marginal changes are truly present.
Grade 2 — Minimal. Definite osteophytes with possible narrowing of the joint space. This is often the grade at which patients first present with mechanical symptoms — morning stiffness, crepitus on flexion, and activity-related pain.
Grade 3 — Moderate. Moderate multiple osteophytes, definite narrowing of joint space, some subchondral sclerosis, and possible deformity of bone contour. This grade typically correlates with consistent pain and functional limitation.
Grade 4 — Severe. Large osteophytes, marked narrowing of joint space, severe subchondral sclerosis, and definite deformity of bone contour. At this stage, the joint space may be completely obliterated on the AP view, and bone-on-bone contact is often present.
Key Radiographic Features
Osteophytes — bony outgrowths at the joint margins — are the earliest and most reliable radiographic sign of OA. Marginal osteophytes at the tibial plateau and femoral condyles often precede measurable joint space loss.
Joint space narrowing reflects cartilage loss and is the primary marker used to track OA progression. Weight-bearing films are essential for accurate measurement; non-weight-bearing films can underestimate narrowing by 30–50%.
Subchondral sclerosis appears as increased bone density beneath the articular surface, reflecting repetitive microtrauma and bone remodelling in response to abnormal load distribution.
Subchondral cysts (geodes) develop in advanced disease as synovial fluid intrudes through microfractures in the sclerotic subchondral bone. Their presence generally indicates Grade 3 or higher.
Clinical Challenges: Inter-Observer Variability
One of the most significant limitations of KL grading is inter-observer variability. Studies have shown moderate agreement (weighted kappa 0.50–0.65) between experienced musculoskeletal radiologists, with the greatest disagreement occurring at the Grade 1–2 boundary.
This variability has real clinical consequences: a patient graded KL-2 by one physician might be graded KL-1 or KL-3 by another, potentially altering the treatment plan — from conservative management to surgical referral.
AI-Assisted KL Grading
Deep learning models trained on large radiograph datasets can classify KL grades with accuracy comparable to experienced radiologists. More importantly, they provide perfect intra-observer consistency — the same image always receives the same grade.
At Salnus, we developed our KL grading model through 21 systematic experiments across 11 deep learning architectures. Our DenseNet-121-based model achieves 84.1% binary OA detection accuracy while producing GradCAM heatmaps that highlight the anatomical regions driving each prediction. This allows surgeons to verify that the AI is focusing on clinically relevant features — joint space, osteophytes, subchondral bone — rather than artefacts.
AI grading also has the advantage of being able to provide quantitative measurements (joint space width in millimetres, angle measurements) alongside the categorical grade, giving surgeons more granular data for treatment decisions.
However, AI is not a replacement for clinical judgement. The KL grade is one data point in a complex decision-making process that includes patient symptoms, functional status, comorbidities, and shared decision-making. AI tools like ours are designed to support the surgeon, not supplant them.
From Research to Clinical Application
Our published research in The Orthopaedic Journal of Sports Medicine (OJSM) on 3D-printed patient-specific guides for knee reconstruction demonstrates Salnus's commitment to bridging the gap between academic research and surgical practice. The KL grading AI module follows the same philosophy: rigorous validation before clinical deployment.
If you are an orthopaedic surgeon interested in evaluating our AI-assisted OA analysis tools, we invite you to explore the Salnus Surgeon Portal or contact our team to discuss a pilot collaboration.
References
- Kellgren JH, Lawrence JS. Radiological assessment of osteo-arthrosis. Ann Rheum Dis. 1957;16(4):494-502.
- Kohn MD, Sassoon AA, Fernando ND. Classifications in brief: Kellgren-Lawrence classification of osteoarthritis. Clin Orthop Relat Res. 2016;474(8):1886-1893.
- Tiulpin A, et al. Automatic knee osteoarthritis diagnosis from plain radiographs: a deep learning-based approach. Sci Rep. 2018;8(1):1727.
Disclaimer: This article is for educational purposes only and does not constitute medical advice. Clinical decisions should be made by qualified healthcare professionals based on individual patient assessment.
Reviewed by the Salnus biomedical engineering team.