Radiomics analysis and classification of esophageal involvement in systemic sclerosis using chest CT images

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Esophageal involvement is a frequent early sign of systemic sclerosis (SSc).
• Current evaluation methods like esophageal manometry are invasive and lack morphological detail.
• There is a need for noninvasive techniques to assess esophageal SSc involvement.

Purpose of the Study:

• To identify CT-based radiomics features associated with esophageal involvement in SSc.
• To assess the accuracy of classifying SSc-related esophageal involvement using these radiomics features.

Main Methods:

• Analysis of chest CT scans from 30 SSc patients and 34 healthy controls.
• Automated esophageal segmentation and radiomics feature extraction from axial CT slices.
• Feature selection and classification using XGBoost algorithm.

Main Results:

• First-order statistical radiomics features at thoracic vertebral levels T3-T5 were significantly linked to esophageal involvement.
• High classification accuracy was achieved using only five to six selected radiomics features.
• The presence or absence of esophageal involvement can be accurately identified with a limited feature set.

Conclusions:

• CT-based radiomics offers a potentially valuable noninvasive method for evaluating esophageal involvement in SSc.
• Further research incorporating pathological data and surrounding tissue analysis could enhance clinical utility.