Should artificial intelligence have lower acceptable error rates than humans?

The first patient was misclassified in the diagnostic conclusion according to a local clinical expert opinion in a new clinical implementation of a knee osteoarthritis artificial intelligence (AI) algorithm at Bispebjerg-Frederiksberg University Hospital, Copenhagen, Denmark. In preparation for the evaluation of the AI algorithm, the implementation team collaborated with internal and external partners to plan workflows, and the algorithm was externally validated. After the misclassification, the team was left wondering: what is an acceptable error rate for a low-risk AI diagnostic algorithm? A survey among employees at the Department of Radiology showed significantly lower acceptable error rates for AI (6.8 %) than humans (11.3 %). A general mistrust of AI could cause the discrepancy in acceptable errors. AI may have the disadvantage of limited social capital and likeability compared to human co-workers, and therefore, less potential for forgiveness. Future AI development and implementation require further investigation of the fear of AI's unknown errors to enhance the trustworthiness of perceiving AI as a co-worker. Benchmark tools, transparency, and explainability are also needed to evaluate AI algorithms in clinical implementations to ensure acceptable performance.

Chronic rhinosinusitis in COPD: A prevalent but unrecognized comorbidity impacting health related quality of life.

INTRODUCTION: Unified airway disease where upper respiratory tract inflammation including chronic rhinosinusitis (CRS) affects lower airway disease is known from asthma, bronchiectasis, cystic fibrosis and primary ciliary dyskinesia but little is known about CRS and health related quality of life in COPD. We investigate firstly, the prevalence of CRS in COPD. Secondly the impact of CRS on HRQoL. Thirdly, risk factors for CRS in COPD. METHODS: cross-sectional study of CRS in 222 COPD patients from 2017 to 2019 according to EPOS2012/2020 and GOLD2019 criteria. Patients completed the COPD assessment test (CAT), Medical Research Council dyspnea scale and Sinonasal outcome test 22 (SNOT22) and questions on CRS symptoms. They then had a physical examination including flexible nasal endoscopy, CT-sinus scan and HRCT-thorax. RESULTS: 22.5% of COPD patients had CRS and 82% of these were undiagnosed prior to the study. HRQoL (CAT, SNOT22 and the SNOT22-nasal symptom subscore) was significantly worse in COPD patients with CRS compared with those without CRS and healthy controls. Multiple logistic regression analysis suggests that the most likely candidate for having CRS was a male COPD patient who actively smoked, took inhaled steroids, had a high CAT and SNOT22_nasal symptom subscore. DISCUSSION: the largest clinical study of CRS in COPD and the only study diagnosing CRS according to EPOS and GOLD. This study supports unified airway disease in COPD. The SNOT22_nasal symptoms subscore is recommended as a standard questionnaire for COPD patients and patients at risk should be referred to an otorhinolaryngologist.

Reliability of standing weight-bearing (0.25T) MR imaging findings and positional changes in the lumbar spine.

OBJECTIVE: To test the reliability and absolute agreement of common degenerative findings in standing positional magnetic resonance imaging (pMRI). METHODS AND MATERIALS: Low back pain patients with and without sciatica were consecutively enrolled to undergo a supine and standing pMRI. Three readers independently evaluated the standing pMRI for herniation, spinal stenosis, spondylolisthesis, HIZ lesions and facet joint effusion. The evaluation included a semi-quantitative grading of spinal stenosis, foraminal stenosis and spinal nerve root compression. The standing pMRI images were evaluated with full access to supine MRI. In case lower grades or the degenerative findings were not present in the supine images, this was reported separately as position-dependent changes. A subsample of 20 pMRI examinations was reevaluated after two months. The reproducibility was assessed by inter- and intra-reader reliability (kappa statistic) and absolute agreement between readers. RESULTS: Fifty-six patients were included in this study. There was fair-to-substantial inter-reader reliability (κ 0.47 to 0.82) and high absolute agreement (72.3% to 99.1%) for the pMRI findings. The intra-reader assessment showed similar reliability and agreement (κ 0.36 to 0.85; absolute agreement: 62.5% to 98.8%). Positional changes between the supine and standing position showed a fair-to-moderate inter- and intra-reader reliability (κ 0.25 to 0.52; absolute agreement: 97.0% to 99.1). CONCLUSION: Evaluation of the lumbar spine for degenerative findings by standing pMRI has acceptable reproducibility; however, positional changes from the supine to the standing position as an independent outcome should be interpreted with caution because of lower reliability, which calls for further standardisation.