Impact of Different Artificial Intelligence User Interfaces on Lung Nodule and Mass Detection on Chest Radiographs.

Purpose: To explore the impact of different user interfaces (UIs) for artificial intelligence (AI) outputs on radiologist performance and user preference in detecting lung nodules and masses on chest radiographs. Materials and Methods: A retrospective paired-reader study with a 4-week washout period was used to evaluate three different AI UIs compared with no AI output. Ten radiologists (eight radiology attending physicians and two trainees) evaluated 140 chest radiographs (81 with histologically confirmed nodules and 59 confirmed as normal with CT), with either no AI or one of three UI outputs: (a) text-only, (b) combined AI confidence score and text, or (c) combined text, AI confidence score, and image overlay. Areas under the receiver operating characteristic curve were calculated to compare radiologist diagnostic performance with each UI with their diagnostic performance without AI. Radiologists reported their UI preference. Results: The area under the receiver operating characteristic curve improved when radiologists used the text-only output compared with no AI (0.87 vs 0.82; P < .001). There was no difference in performance for the combined text and AI confidence score output compared with no AI (0.77 vs 0.82; P = .46) and for the combined text, AI confidence score, and image overlay output compared with no AI (0.80 vs 0.82; P = .66). Eight of the 10 radiologists (80%) preferred the combined text, AI confidence score, and image overlay output over the other two interfaces. Conclusion: Text-only UI output significantly improved radiologist performance compared with no AI in the detection of lung nodules and masses on chest radiographs, but user preference did not correspond with user performance.Keywords: Artificial Intelligence, Chest Radiograph, Conventional Radiography, Lung Nodule, Mass Detection© RSNA, 2023.

Fat fraction quantification of bone marrow in the lumbar spine using the LiverLab assessment tool in healthy adult volunteers and patients with Gaucher disease.

BACKGROUND: Magnetic Resonance Imaging is used for evaluation of bone in Gaucher disease (GD), but a widely available quantitative scoring method remains elusive. AIMS: The study purpose was to assess the reproducibility of the LiverLab tool for assessing bone marrow fat fraction (FF) and determine whether it could differentiate GD patients from healthy subjects. METHODS: Ten healthy volunteers and 18 GD patients were prospectively recruited. FF was calculated at L3, L4 and L5. GD patient bone marrow burden (BMB) score assessed by one observer. Inter and intra-rater agreement assessed with Bland-Altman data plots. Differences in FF between healthy volunteers versus GD patients and between subjects treated versus not treated assessed using two-sample t-tests. In GD patients, the relationship between FF, BMB and glucosylsphingosine was determined using the Pearson's correlation coefficient. RESULTS: Healthy volunteer mean FF was 0.36, standard deviation (SD) 0.10 (range 0.20-0.57). Intra and inter-rater SD were both 0.02. GD patient mean FF was 0.40, SD 0.13 (range 0.09-0.57). No statistical difference was shown between healthy volunteers and GD patients (P = 0.447) or between GD patients whether on enzyme replacement therapy or not (P = 0.090). No significant correlation between mean FF and total BMB (r = -0.525, P = 0.253) or between FF and glucosylsphingosine levels (r = 0.287, P = 0.248). CONCLUSION: Excellent reproducibility of LiverLab FF measurements across studies and observers is comparable to Dixon quantitative chemical shift imaging (QCSI). Lack of statistical difference between GD patients and controls may be explained by limited patient numbers, active treatment or mild disease severity in untreated patients.

Intraobserver and interobserver variability of the bone marrow burden (BMB) score for the assessment of disease severity in Gaucher disease. Possible impact of reporting experience.

AIM: To evaluate the intraobserver and interobserver agreement for bone marrow burden (BMB) scores for individual examinations and for the change in BMB score over time in the same patient. METHODS: A total of 119 sets of MR images of the lumbar spine and femora from 60 patients with Gaucher disease were included. Each set of MR images was scored using the BMB score independently by two experienced MSK radiologists. One radiologist performed a second read four weeks later. Intraobserver and interobserver agreement was assessed using Bland-Altman analysis and weighted kappa scores. RESULTS: BMB scores (n=119) demonstrated fair intraobserver agreement (weighted kappa=0.53) with a mean difference of -0.20 and 95% limits of agreement (LOA) of (-3.41, 3.01). Inter observer agreement was poor with weighted kappa 0.28 with mean difference of -0.16 and 95% LOA of (-4.45, 4.11). Change in BMB scores over time (n=59) demonstrated poor/fair intraobserver agreement (weighted kappa 0.41, mean difference-0.20 and 95% LOA (-4.35, 3.94)). Interobserver agreement was poor (weighted kappa 0.25, mean difference -0.12 with wide 95% LOA (-6.23, 5.99)). CONCLUSION: Significant interobserver, and to a lesser extent intraobserver, variation occurs with blinded BMB scoring of Gaucher disease.

Prospective observational study of physical functioning, physical activity, and time outdoors and the risk of hip fracture: a population-based cohort study of 158,057 older adults in the 45 and up study.

Low levels of physical activity or sun exposure and limitations to physical functioning (or disability) have been identified as possible risk factors for hip fracture. However, these factors are closely related, and data on their independent and joint association with risk of hip fracture are limited. A total of 158,057 individuals aged ≥45 years sampled from the general population of New South Wales, Australia, from the prospective 45 and Up Study completed a baseline postal questionnaire in 2006 to 2009 including data on physical activity (Active Australia questionnaire); sun exposure (usual time outdoors); and physical functioning (Medical Outcomes Score-Physical Functioning; scored 0 to 100). Incident first hip fractures were ascertained by linkage to administrative hospital data (n = 293; average follow-up 2.3 years). The relative risk (RR) of hip fracture was estimated using Cox proportional hazards. Poorer physical functioning, lower physical activity, and less time outdoors were positively related to each other at baseline and individually associated with significantly increased hip fracture risk. However, physical activity and time outdoors were not significantly related to hip fracture risk after adjustment for baseline physical functioning or when analysis was restricted to those with no or mild baseline physical limitation. In contrast, physical functioning remained strongly related to hip fracture risk after adjustment for the other two factors; compared with the group without limitation (100), the RR of hip fracture among those with mild (75-95), moderate (50-70), severe (25-45), and greatest (0-20) level of physical limitation was 1.38 (95% confidence interval [CI] 0.88-2.14), 2.14 (1.29-3.53), 3.87 (2.31-6.44), and 5.61 (3.33-9.42), respectively. The findings suggest that limitation in physical functioning, but not physical activity or time outdoors, is strongly related to hip fracture risk. The apparent increased risk of hip fracture previously described for low physical activity or sun exposure may be, at least in part due to uncontrolled confounding.

Hip fracture risk in relation to vitamin D supplementation and serum 25-hydroxyvitamin D levels: a systematic review and meta-analysis of randomised controlled trials and observational studies.

BACKGROUND: Vitamin D supplementation for fracture prevention is widespread despite conflicting interpretation of relevant randomised controlled trial (RCT) evidence. This study summarises quantitatively the current evidence from RCTs and observational studies regarding vitamin D, parathyroid hormone (PTH) and hip fracture risk. METHODS: We undertook separate meta-analyses of RCTs examining vitamin D supplementation and hip fracture, and observational studies of serum vitamin D status (25-hydroxyvitamin D (25(OH)D) level), PTH and hip fracture. Results from RCTs were combined using the reported hazard ratios/relative risks (RR). Results from case-control studies were combined using the ratio of 25(OH)D and PTH measurements of hip fracture cases compared with controls. Original published studies of vitamin D, PTH and hip fracture were identified through PubMed and Web of Science databases, searches of reference lists and forward citations of key papers. RESULTS: The seven eligible RCTs identified showed no significant difference in hip fracture risk in those randomised to cholecalciferol or ergocalciferol supplementation versus placebo/control (RR = 1.13[95%CI 0.98-1.29]; 801 cases), with no significant difference between trials of <800 IU/day and > or = 800 IU/day. The 17 identified case-control studies found 33% lower serum 25(OH)D levels in cases compared to controls, based on 1903 cases. This difference was significantly greater in studies with population-based compared to hospital-based controls (chi(2)(1) (heterogeneity) = 51.02, p < 0.001) and significant heterogeneity was present overall (chi(2)(16) (heterogeneity) = 137.9, p < 0.001). Serum PTH levels in hip fracture cases did not differ significantly from controls, based on ten case-control studies with 905 cases (chi(2)(9) (heterogeneity) = 149.68, p < 0.001). CONCLUSIONS: Neither higher nor lower dose vitamin D supplementation prevented hip fracture. Randomised and observational data on vitamin D and hip fracture appear to differ. The reason for this is unclear; one possible explanation is uncontrolled confounding in observational studies. Post-fracture PTH levels are unrelated to hip fracture risk.

Assessing vitamin D status: pitfalls for the unwary.

The use of vitamin D testing has grown rapidly in the recent times as a result of increased interest in the role of vitamin D in health. Although the generally accepted measure of vitamin D status is circulating 25(OH)D concentration, there is little consensus on which assay method should be used. Commonly used assays include competitive protein-binding assay, RIA, enzyme immunoassay, chemiluminescence immunoassays, HPLC, and LC-MS/MS, each with its own advantages and disadvantages. However, there is significant interassay and interlaboratory variability in measurements. Our simulation of the published data showed that using a deficiency cut-point of 50 nmol/L, 57% of samples assessed using a chemiluminescence immunoassay were classified as deficient compared with 41% of samples assessed using LC-MS/MS; a 20% misclassification rate. Similar rates of misclassification were seen at 75 nmol/L. This has implications for clinical practice and decision limits for vitamin D supplementation, suggesting that cut-points should be assay specific rather than universal and that greater harmonization between laboratories is required. Newer assays using alternative biological samples to determine the circulating 25(OH)D have been proposed and advances in the genetics of vitamin D and the role of vitamin D-binding protein may improve future assay accuracy.