Interpretable machine learning for severity classification of thyroid eye disease using orbital anatomical features - Takeaways - MDSpire

Interpretable machine learning for severity classification of thyroid eye disease using orbital anatomical features

  • By

  • Ruixin Shi

  • Leiming Gao

  • Shengzhi Jiao

  • Liuzi Wang

  • Jianing Li

  • Bei Wang

  • June 19, 2026

  • 0 min

Share

  • 1

    Thyroid eye disease (TED) severity assessment using EUGOGO is subjective and prone to interobserver variability.

  • 2

    MRI-derived anatomical measurements, such as ocular protrusion, offer objective features for TED severity classification.

  • 3

    Random Forest with class weighting achieved the highest AUC of 0.811 in class-imbalance strategies for TED severity classification.

  • 4

    Feature importance analysis identified ocular protrusion as the top predictor of TED severity, followed by rectus muscle thickness.

  • 5

    Controlling for longitudinal redundancy and intra-patient correlations is crucial for accurate model evaluation and generalizability.

Original Source(s)

Interpretable machine learning for severity classification of thyroid eye disease using orbital anatomical features

  • by Ruixin Shi, Leiming Gao, Shengzhi Jiao, Liuzi Wang, Jianing Li, Bei Wang

  • June 19, 2026

Related Content

Taking FLIGHT

ViaLase announces first commercial FLIGHT procedures in Europe

  • June 18, 2026

  • 2 min
The traditional Chinese medicine bloodletting therapy at the Taiyang acupoint (No.EX-HN5) for the treatment of periorbital venous malformation: a case report

Confidence in managing open-globe injuries and endophthalmitis: a pre- and postwet-lab training evaluation