Multimodal imaging-based interpretable radiomics for differentiating brucella and tuberculosis spondylitis: a...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Infectious spondylitis, including tuberculosis spondylitis (TS) and brucella spondylitis (BS), poses diagnostic challenges due to overlapping clinical presentations.
• Accurate differentiation is vital for effective treatment and improved patient outcomes.

Purpose of the Study:

• To evaluate the potential of multimodal imaging-based radiomics for differentiating between tuberculosis spondylitis (TS) and brucella spondylitis (BS).

Main Methods:

• Retrospective cohort study of 195 patients (TS/BS) with external validation (57 patients).
• Standardized MRI (T1WI, T2WI, FS T2WI) and CT imaging protocols were used.
• Radiomics features were extracted, and nine predictive models were developed and validated using cross-validation.

Main Results:

• Multi-modality radiomics models combining CT, T1WI, T2WI, and FS T2WI features achieved an optimal Area Under the Curve (AUC) of 0.8136 in the validation dataset.
• Model performance was confirmed via ROC analysis, decision curve analysis, and calibration plots.
• SHAP analysis identified key features influencing model predictions.

Conclusions:

• Multimodal imaging-based radiomics demonstrates significant potential for the differential diagnosis of brucella spondylitis and tuberculosis spondylitis.
• This approach can enhance diagnostic precision for challenging spinal infections.