Glenoid bone loss in anterior shoulder dislocation: a multicentric study to assess the most reliable imaging method.

PURPOSE: The aim of this multicentric study was to assess which imaging method has the best inter-reader agreement for glenoid bone loss quantification in anterior shoulder instability. A further aim was to calculate the inter-method agreement comparing bilateral CT with unilateral CT and MR arthrography (MRA) with CT measurements. Finally, calculations were carried out to find the least time-consuming method. METHOD: A retrospective evaluation was performed by 9 readers (or pairs of readers) on a consecutive series of 110 patients with MRA and bilateral shoulder CT. Each reader was asked to calculate the glenoid bone loss of all patients using the following methods: best fit circle area on both MRA and CT images, maximum transverse glenoid width on MRA and CT, CT PICO technique, ratio of the maximum glenoid width to height on MRA and CT, and length of flattening of the anterior glenoid curvature on MRA and CT. Using Pearson's correlation coefficient (PCC), the following agreement values were calculated: the inter-reader for each method, the inter-method for MRA with CT quantifications and the inter-method for CT best-fit circle area and CT PICO. Statistical analysis was carried out to compare the time employed by the readers for each method. RESULTS: Inter-reader agreement PCC mean values were the following: 0.70 for MRA and 0.77 for CT using best fit circle diameter, 0.68 for MRA and 0.72 for CT using best fit circle area, 0.75 for CT PICO, 0.64 for MRA and 0.62 for CT anterior straight line and 0.49 for MRA and 0.43 for CT using length-to-width ratio. CT-MRA inter-modality PCC mean values were 0.9 for best fit circle diameter, 0.9 for best fit circle area, 0.62 for anterior straight line and 0.94 for length-to-width methods. PCC mean value comparing unilateral CT with PICO CT methods was 0.8. MRA best fit circle area method was significantly faster than the same method performed on CT (p = 0.031), while no significant difference was seen between CT and MRA for remaining measurements. CONCLUSIONS: CT PICO is the most reliable imaging method, but both CT and MRA can be reliably used to assess glenoid bone loss. Best fit circle area CT and MRA methods are valuable alternative measurement techniques.

Comparative analysis of explainable machine learning prediction models for hospital mortality.

BACKGROUND: Machine learning (ML) holds the promise of becoming an essential tool for utilising the increasing amount of clinical data available for analysis and clinical decision support. However, the lack of trust in the models has limited the acceptance of this technology in healthcare. This mistrust is often credited to the shortage of model explainability and interpretability, where the relationship between the input and output of the models is unclear. Improving trust requires the development of more transparent ML methods. METHODS: In this paper, we use the publicly available eICU database to construct a number of ML models before examining their internal behaviour with SHapley Additive exPlanations (SHAP) values. Our four models predicted hospital mortality in ICU patients using a selection of the same features used to calculate the APACHE IV score and were based on random forest, logistic regression, naive Bayes, and adaptive boosting algorithms. RESULTS: The results showed the models had similar discriminative abilities and mostly agreed on feature importance while calibration and impact of individual features differed considerably and did in multiple cases not correspond to common medical theory. CONCLUSIONS: We already know that ML models treat data differently depending on the underlying algorithm. Our comparative analysis visualises implications of these differences and their importance in a healthcare setting. SHAP value analysis is a promising method for incorporating explainability in model development and usage and might yield better and more trustworthy ML models in the future.