Machine Learning May Be an Alternative to BIPSS in the Differential Diagnosis of ACTH-dependent Cushing Syndrome
Authors' conclusion
ML algorithms have the potential to serve as an alternative decision-support tool to invasive procedures in the differential diagnosis of ACTH-dependent CS.
Study context
- Population
- Patients with ACTH-dependent Cushing syndrome (CS); features: biochemical and radiological
- Participants
- 106 patients
- Comparison
- Machine-learning (ML) algorithm for classifying ACTH-dependent CS into Cushing disease (CD) vs ectopic ACTH syndrome (EAS)
- CDn = 80
- EASn = 26
- Outcome
- AUROC of the logistic regression model for classifying CD vs EAS
Methodological detailsmachine learning, logistic regression, cross-validation, Algorithm made available to clinicians via a user-friendly interface
- Machine learning
- Logistic regression
- Cross-validation
- Algorithm made available to clinicians via a user-friendly interface
Performance
As reported in the abstract
Classification of ACTH-dependent CS into CD and EAS (logistic regression ML model)
Average AUROC in cross-validation
0.850
Diagnostic accuracy
86%
Further findings
Most important model determinants (SHAP values)
- 2-day 2-mg dexamethasone suppression test, >50% suppression in 8-mg high-dose dexamethasone test, late-night salivary cortisol, and pituitary adenoma diameter were the most important determinants
Original publication
Machine Learning May Be an Alternative to BIPSS in the Differential Diagnosis of ACTH-dependent Cushing Syndrome. J Clin Endocrinol Metab. Published 19 March 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →