Multiparametric Lung MRI With Diffusion-Weighted Imaging in Lung-RADS 4 Lesion Characterization

OBJECTIVES

Aim 1. Determine the sensitivity and specificity of multiparametric chest MRI with DWI in differentiating malignant from benign Lung-RADS 4 lesions.

Aim 2. Determine whether DWI alone is as effective as the full chest protocol in malignancy classification of Lung-RADS 4 lesions.

Aim 3. Compare the diagnostic performance of chest MRI to PET/CT in the characterization of Lung-RADS 4 lesions.

DESIGN

Single-center prospective observational cohort study of diagnostic accuracy at the University of Florida. Participants identified from the institutional lung cancer screening program with a Lung-RADS 4 (4A, 4B, 4X) finding on standard-of-care low-dose CT undergo one non-contrast research chest MRI. There is no randomization and no comparator group; the comparator is the reference standard (final clinical diagnosis) and, for Aim 3, the participant's standard-of-care PET/CT where performed.

INDEX TEST

All examinations are performed on a 3.0 Tesla scanner (GE Healthcare or Siemens Healthineers) with a vendor-specific surface coil; an interlocking neck coil may be added to the torso coil for apical lesions. No intravenous or other contrast material is administered. Sequences comprise axial breath-hold 2D in- and opposed-phase gradient echo T1-weighted imaging; axial breath-hold and/or respiratory-triggered DWI using multi-shot fat-suppressed echo-planar imaging with two b values; DWI sequences with restricted z-axis coverage centered on the nodule (EPI-DWI, FOCUS/ZOOMit-EPI, MUSE/RESOLVE); and axial respiratory-triggered T2-weighted imaging with fat suppression or Dixon. Ultra-short echo time imaging and T1/T2 mapping centered on the nodule may be added if time allows. Pulse-sequence parameters are optimized to the vendor platform during a pre-clinical optimization phase and after the first participants are scanned.

IMAGE ANALYSIS

Anonymized MRI examinations are read independently by two cardiothoracic radiologists with more than 5 years of MRI experience. Readers have the companion low-dose CT but are blinded to clinical data and to any additional imaging. Reading proceeds in two rounds: DWI alone in conjunction with the LDCT (Aim 2), then the full set of sequences (Aim 1). Each examination is classified as (1) negative / low suspicion for malignancy, (2) indeterminate, would benefit from short-term follow-up, or (3) positive / high suspicion for malignancy. Quantitative measures are the apparent diffusion coefficient from a circular or elliptical ROI over the lowest-signal component of the nodule, and the nodule-to-spinal-cord signal intensity ratio measured at the same level.

REFERENCE STANDARD

The final etiology of the lesion is determined by histopathological and/or microbiological data when available, or otherwise by imaging and clinical follow-up for at least 24 months. A lesion that disappears or decreases in size on follow-up is deemed benign; a lesion that remains stable for at least 24 months is deemed benign.

STATISTICAL ANALYSIS

Sensitivity and specificity of chest MRI and of DW-MRI alone are reported with 95% confidence intervals, computed both excluding and including indeterminate examinations. For participants whose standard-of-care PET/CT was performed within a 3-week interval, sensitivity and specificity of the two modalities are compared pairwise by McNemar's test. Thresholds for ADC, lesion-to-spinal-cord ratio, native T1 and T2 are explored by ROC analysis at Youden's index. Interobserver agreement is assessed by kappa for categorical variables and intraclass correlation for continuous variables. Analyses are performed in R version 4.4.1 or higher at a significance level of 0.05. No interim analysis is planned.