The Incremental Value of Native T1 Mapping-Derived Radiomics for The Diagnosis of Amyloid Light-Chain Cardiac Amyloidosis

Abstract

Rationale and Objectives

This study aimed to assess the incremental diagnostic value of non-contrast T1 mapping-derived radiomics in patients with amyloid light-chain cardiac amyloidosis (AL-CA).

Methods

We retrospectively collected 86 patients with suspected AL-CA and 28 control patients who underwent cardiac MRI at 3.0 T in our institution, and the MRI data were divided into a training set and a test set. Radiomic features were extracted based on native T1 maps using a freely available software package. We applied LASSO logistic regression method to select radiomic features with high diagnostic value of AL-CA and develop a predictive model. The diagnostic performance of the radiomics model was evaluated using receiver operating characteristic curve analysis and compared to T1 values.

Results

A total of 70 people were diagnosed with AL-CA, and cardiac involvement was observed in 202 myocardial slicers. The accuracy of T1 values for the diagnosis of myocardial involvement was 0.886, with a threshold value of 1375 ms. The radiomics score comprised a total of three features. The radiomics score demonstrated significantly higher sensitivity in detecting myocardial involvement compared to T1 values in both the training set (AUC 0.886 vs. 0.924, P = 0.025) and the test set (0.862 vs 0.915, P = 0.026). The combined model comprising T1 values and a radiomic feature named S(4,−4) Correlat showed higher diagnostic performance in comparison to T1 values alone both in the training and test sets, with AUC values of 0.929 and 0.909, respectively.

Conclusion

The radiomic features derived from native T1 mapping demonstrated incremental value for the diagnosis of AL-CA, which may be an alternative to T1-derived ECV to avoid the use of contrast in patients with suspected myocardial involvement in systemic amyloidosis.