KOALA

The KOALA software is an automated image processing system that utilizes machine learning algorithms to analyze CR or DX knee radiographs for signs of osteoarthritis by measuring joint space width and detecting features based on OARSI and Kellgren-Lawrence criteria. Its purpose is to provide quantitative and qualitative data to aid trained medical professionals in the assessment and grading of knee OA, supported by clinical validation demonstrating its accuracy in feature detection and measurement.

IB Lab's KOALA is a fully-automated radiological image processing software designed to assist medical professionals in evaluating knee osteoarthritis (OA) from computed (CR) or directly digital (DX) X-ray images. It employs computer vision and machine learning algorithms to analyze posterior-anterior or anterior-posterior knee radiographs of adult patients. The software provides quantitative measurements of minimum joint space width and indicators for the presence or absence of key radiographic OA features, including sclerosis, joint space narrowing, and osteophytes, based on OARSI criteria, as well as an assessment of radiographic knee OA according to Kellgren & Lawrence Grading. These outputs are compiled into a KOALA report, accessible via any DICOM viewer workstation, and are intended for use by trained professionals such as radiologists, orthopedists, physicians, and medical technicians, though not as a sole basis for diagnosis. The performance of KOALA was validated using a substantial dataset of 6597 radiographs from 1149 individuals in the Osteoarthritis Initiative (OAI) study, with ground truth established by three physicians. For OA feature detection, KOALA demonstrated sensitivities and specificities as follows: Kellgren-Lawrence status (KL ≥ 2) showed 0.87 sensitivity and 0.83 specificity; joint space narrowing (OARSI > 0) had 0.83 sensitivity and 0.80 specificity; osteophytosis (OARSI > 0) achieved 0.86 sensitivity and 0.79 specificity; and sclerosis (OARSI > 0) resulted in 0.82 sensitivity and 0.80 specificity. Furthermore, joint space width (JSW) measurements showed good agreement with reference measurements from the OAI study, with orthogonal linear regression yielding a slope of 1.02 and intercept of -0.08 mm for medial JSW, and a slope of 0.97 and intercept of 0.08 mm for lateral JSW, supporting the device's reliability.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does KOALA's performance in detecting early-stage osteoarthritis compare to its performance in more advanced stages, and are there specific thresholds for OARSI or Kellgren-Lawrence grades where its accuracy is maximized?; What impact does variability in X-ray acquisition parameters (e.g., beam angle, patient positioning consistency for fixed-flexion views) have on the robustness and reliability of KOALA's joint space width measurements and feature detection?; Could the KOALA software be adapted or expanded to analyze longitudinal changes in knee OA progression in individual patients over time, potentially serving as a tool for monitoring treatment efficacy?; What were the specific adjudication procedures used by the three physicians to establish ground truth for the OAI dataset, and how was inter-observer variability among them managed before creating the reference standard?; Beyond the reported sensitivity and specificity, what are the positive predictive value (PPV) and negative predictive value (NPV) of KOALA's indicators in a typical clinical population with varying OA prevalence?