AI in Process Automation

Machine-learning models for diagnosis of rotator cuff tears in osteoporosis patients based on anteroposterior X-rays of the shoulder joint

OBJECTIVE

This study aims to diagnose Rotator Cuff Tears (RCT) and classify the severity of RCT in patients with Osteoporosis (OP) through the analysis of shoulder joint anteroposterior (AP) X-ray-based localized proximal humeral bone mineral density (BMD) measurements and clinical information based on machine learning (ML) models.

METHODS

A retrospective cohort of 89 patients was analyzed, including 63 with both OP and RCT (OPRCT) and 26 with OP only. The study analyzed a series of shoulder radiographs from April 2021 to April 2023. Grayscale values were measured after plotting ROIs based on AP X-rays of shoulder joint. Five kinds of ML models were developed and compared based on their performance in predicting the occurrence and severity of RCT from ROIs' greyscale values and clinical information (age, gender, advantage side, lumbar BMD, and acromion morphology (AM)). Further analysis using SHAP values illustrated the significant impact of selected features on model predictions.

RESULTS

R1-6 had a positive correlation with BMD respectively. The nine variables, including greyscale R1-6, age, BMD, and AM, were used in the prediction models. The RF model was determined to be superior in effectively diagnosing RCT in OP patients, with high AUC scores of 0.998, 0.889, and 0.95 in the training, validation, and testing sets, respectively. SHAP values revealed that the most influential factors on the diagnostic outcomes were the grayscale values of all cancellous bones in ROIs. A column-line graph prediction model based on nine variables was constructed, and DCA curves indicated that RCT prediction in OP patients was favored based on this model. Furthermore, the RF model was also the most superior in predicting the types of RCT within the OPRCT group, with an accuracy of 86.364% and 73.684% in the training and test sets, respectively. SHAP values indicated that the most significant factor affecting the predictive outcomes was the AM, followed by the grayscale values of the greater tubercle, among others.

CONCLUSIONS

ML models, particularly the RF algorithm, show significant promise in diagnosing RCT occurrence and severity in OP patients using conventional shoulder X-rays based on the nine variables. This method presents a cost-effective, accessible, and non-invasive diagnostic strategy that has the potential to substantially enhance the early detection and management of RCT in OP patient population.

Integration of artificial intelligence and plasma steroidomics with laboratory information management systems: application to primary aldosteronism

Objectives

Mass spectrometry-based steroidomics combined with machine learning (ML) provides a potentially powerful approach in endocrine diagnostics, but is hampered by limitations in the conveyance of results and interpretations to clinicians. We address this shortcoming by integration of the two technologies with a laboratory information management systems (LIMS) model.

Methods

The approach involves integration of ML algorithm-derived models with commercially available mathematical programming software and a web-based LIMS prototype. To illustrate clinical utility, the process was applied to plasma steroidomics data from 22 patients tested for primary aldosteronism (PA).

Results

Once mass spectrometry data are uploaded into the system, automated processes enable generation of interpretations of steroid profiles from ML models. Generated reports include plasma concentrations of steroids in relation to age- and sex-specific reference intervals along with results of ML models and narrative interpretations that cover probabilities of PA. If PA is predicted, reports include probabilities of unilateral disease and mutations of KCNJ5 known to be associated with successful outcomes of adrenalectomy. Preliminary results, with no overlap in probabilities of disease among four patients with and 18 without PA and correct classification of all four patients with unilateral PA including three of four with KCNJ5 mutations, illustrate potential utility of the approach to guide diagnosis and subtyping of patients with PA.

Conclusions

The outlined process for integrating plasma steroidomics data and ML with LIMS may facilitate improved diagnostic-decision-making when based on higher-dimensional data otherwise difficult to interpret. The approach is relevant to other diagnostic applications involving ML.