Clinical Scorecard: Machine Learning Expands Across Endocrinology

At a Glance

Key Highlights

• 1,130 studies identified, with 68% focused on thyroid diseases.
• ML models show high diagnostic performance in thyroid nodule evaluation.
• Deep learning reduced unnecessary biopsies by 27% while maintaining accuracy.
• ML applications in adrenal disorders achieved AUC values above 0.94.
• Limitations include lack of model transparency and small sample sizes.

Guideline-Based Recommendations

Diagnosis

• Use ML models for imaging-based evaluations and risk stratification.

Management

• Integrate ML findings into clinical decision-making with specialist supervision.

Monitoring & Follow-up

• Employ ML for predicting postoperative outcomes and complications.

Risks

• Address data imbalance and the need for external validation of ML models.

Patient & Prescribing Data

Individuals with thyroid, pituitary, adrenal, and parathyroid disorders.

ML can enhance diagnostic accuracy and reduce unnecessary interventions.

Clinical Best Practices

• Encourage interdisciplinary collaboration between healthcare professionals and data scientists.
• Focus on high-quality, well-designed ML research in endocrinology.
• Standardize reporting and validation processes for ML models.

References

• Endocrine Review