Differentiation of Pulmonary Lymphoma Manifestations and Nonlymphoma Infiltrates in Possible Invasive Fungal Disease Using Fast T1-weighted Magnetic Resonance Imaging at 3 T Comparison of Texture Analysis, Mapping, and Signal Intensity Quotients.

PURPOSE: This study aimed to evaluate the diagnostic performance of texture analysis (TA), T1 mapping, and signal intensity quotients derived from fast T1-weighted gradient echo (T1w GRE) sequences for differentiating pulmonary lymphoma manifestations and nonlymphoma infiltrates in possible invasive fungal disease in immunocompromised hematological patients.

MATERIALS AND

METHODS

Twenty patients with hematologic malignancies and concomitant immunosuppression (including 10 patients with pulmonary lymphoma manifestations and 10 patients with nonlymphoma infiltrates) prospectively underwent 3 T magnetic resonance imaging using a conventional T1w GRE sequence and a T1w GRE mapping sequence with variable flip angle. A region of interest was placed around the most representative lesion in each patient. TA was performed using PyRadiomics. T1 relaxation times were extracted from precompiled maps and calculated manually. Signal intensity quotients (lesion/muscle) were calculated from conventional T1w GRE sequences.

RESULTS

Of all TA features, variance, mean absolute deviation, robust mean absolute deviation, interquartile range, and minimum were significantly different between the 2 entities (P<0.05), with excellent diagnostic performance in receiver operating characteristic analysis (area under the curve [AUC] >80%). Neither T1 relaxation times from precompiled maps (AUC=63%; P=0.353) nor manual calculation (AUC=63%; P=0.353) nor signal intensity quotients (AUC=70%; P=0.143) yielded significant differences.

CONCLUSIONS

TA from fast T1w GRE images can differentiate pulmonary lymphoma manifestations and nonlymphoma infiltrates in possible invasive fungal disease with excellent diagnostic performance using the TA features variance, mean absolute deviation, robust mean absolute deviation, interquartile range, and minimum. Combining a fast T1w GRE sequence with TA seems to be a promising tool to differentiate these 2 entities noninvasively.