Fully automated image processing to support in analysis and reporting of AP/PA unliateral or bilateral and LAT planar images of the Knee
Fully -automated image processing software device intended to aid medical professionals in the analysis and reporting of weight-bearing AP/PA unilateral or bilateral, and lateral (LAT) planar images of the knee, acquired with a suspicion of osteoarthritis in the tibiofemoral and/or patellofemoral joints.
CR or DX images of the knee
Copy of regions of interest from the original radiographs with overlays that mark the minimum Joint Space Width (JSW) T table with the measurement of the minimum JSW in millimeters (mm), information on the presence or absence of joint space narrowing, osteophytes, and sclerosis based on OARSI gradings, and information on the presence or absence of radiographic knee osteoarthritis (OA) based on the Kellgren-Lawrence (KL) score.
The study included 50 patients (62% female, median age of 68 years, range 20-90) resulting in 99 knees evaluated for osteoarthritis (OA) severity using the Kellgren-Lawrence (KL) grading system. KL 0: Sensitivity 100% (CI: 100-100), Specificity 76% (CI: 68-85)KL 3: Sensitivity 94% (CI: 83-100), Specificity 91% (CI: 83-100)
Overall accuracy ranged from 84% for multiclass accuracy (95% CI: 77-91%) and up to 97.8% in weighted accuracy, indicating strong diagnostic precision. Agreement between the AI tool and radiology consultants was 0.88 (95% CI: 0.82–0.92), demonstrating high reliability close to consultant consensus levels.
Multiclass F1-score and p-values were calculated to measure inter-reader consistency, with significant reliability (p < 0.05) across different KL grades
The AI tool processes bilateral knee radiographs in 8-60 seconds per case depending on the availability of a graphical processing unit (GPU), supporting fast and efficient radiographic evaluation.
The tool’s automated grading of OA severity reduces the need for manual assessment, especially beneficial in clinical settings with a high volume of knee radiographs. It minimizes reading time, particularly for routine evaluations, allowing radiologists to focus on more complex cases.
By automating OA grading, RBknee reduces the need for radiologists in standard knee OA evaluations, resulting in cost savings, especially in resource-constrained environments.
The tool assists radiology departments by allowing non-specialized staff to carry out OA severity grading with reliability comparable to musculoskeletal specialists. This optimizes the allocation of specialized radiologists for more intricate cases.
Accurate and consistent grading can reduce the need for additional imaging or second opinions, lowering the long-term healthcare costs associated with OA management by preventing misdiagnoses and ensuring timely and appropriate care
