Related Experiment Video
Updated: Jul 2, 2026
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
6.6K
CT-Based Machine Learning Radiomics Analysis to Diagnose Dysthyroid Optic Neuropathy
Lan Ma1, Xue Jiang1, Xuan Yang1
1Beijing Tongren Eye Center, Beijing Tongren Hospital, Beijing Ophthalmology and Visual Science Key Lab, Capital Medical University, Beijing, China.
Seminars in Ophthalmology
|February 19, 2025
Summary
This study developed CT-based machine learning radiomics models for diagnosing dysthyroid optic neuropathy (DON). The models showed excellent diagnostic ability, potentially improving convenience for DON diagnosis.
Area of Science:
- Ophthalmology
- Radiology
- Medical Imaging
- Machine Learning
Background:
- Dysthyroid optic neuropathy (DON) is a vision-threatening complication of thyroid-associated ophthalmopathy (TAO).
- Accurate and timely diagnosis of DON is crucial for effective management.
- Current diagnostic methods may have limitations, necessitating advanced imaging techniques.
Purpose of the Study:
- To develop and validate computed tomography (CT)-based machine learning radiomics models for the diagnosis of dysthyroid optic neuropathy (DON).
Main Methods:
- A retrospective study included 57 patients with TAO, with CT scans analyzed for radiomics features.
- Features were extracted using Pyradiomics and selected via LASSO regression.
- Random Forest, SVM, and LR models were trained and validated using ROC curves and AUC.
Main Results:
- Five informative radiomics features were identified using LASSO regression.
- The models achieved AUCs ranging from 0.80 to 0.86 in the test set.
- No significant differences were found between the Random Forest, SVM, and LR models.
Conclusions:
- CT-based machine learning radiomics analysis demonstrates excellent diagnostic capability for DON.
- These models offer a promising tool to enhance the convenience and accuracy of DON diagnosis.
- Further research may explore integration into clinical workflows for improved patient outcomes.
