LAUSR

The initial screening of medical images remains a fundamental bottleneck to the effective diagnosis of time-sensitive diseases in many settings. This has been mitigated in recent years by the introduction of deep learning-based approaches, in many imaging fields. Perhaps one of the most popular applications has been toward replicating the judgment of human experts as closely as possible, with reference to some established scoring standard. This is the approach that Dong et al. have taken, in their study on classifying metacarpophalangeal (MCP) synovial proliferation (SP) in rheumatoid arthritis, from ultrasound images. The relevant scoring system is then the OMERACT-EULAR Synovitis Scoring (OESS) standard, which grades SP in a range of L0– L3. Area under receiver operating characteristic curve of around 0.90 were reported for the clinically relevant binary classification tasks of distinguishing between L0 and non-L0, and L0/L1 versus L2/L3.