Due to increased interest in pulmonary MRI as a radiation free alternative to CT for lung cancer screening, we analyzed MRI characteristics of pulmonary nodules with different non-enhanced sequences.
Several studies have shown the feasibility of MRI for the detection of pulmonary nodules [1-4]. The most predictive parameters of malignancy are nodule size, growth and the presence of spicules [5-8]. For MRI, high signal on T2-weighted images, diffusion restriction, and intense enhancement [2, 9-14] are indicative for lung cancer. The aims of this study were to analyze MRI characteristics of pulmonary nodules with regards to malignancy, and to find out which MRI sequence is most suitable for nodule detection in lung cancer screening.
82 participants of a lung cancer screening (age 50-70 years) were included. MRI datasets of 32 subjects with nodules ≥6 mm on LDCT were anonymized and randomly presented together with datasets of 50 control subjects not showing any nodules or only nodules <4 mm on LDCT. Acquired MRI sequences were T2 short tau inversion recovery (T2-STIR), regular T2 (T2), balanced steady state free precession (bSSFP), T1-weighted high resolution isotropic volume excitation (THRIVE), and diffusion-weighted images (DWI).
Each MRI sequence was
viewed separately and prospectively.
First, the
presence of nodules was assessed. Second, in cases
of positive findings, nodule
size and nodule shape were noted. Third, ROI-based signal intensities were measured inside the lesion with a maximum possible size, and inside the lung with a predefined size to calculate lesion contrast. The ADC-value was
measured, if a lesion showed diffusion restriction. MRI findings were correlated with LDCT as reference.
There were 53 different nodules ≥6 mm in 82 individuals, seven of them being false positive. Lung cancer was revealed in 8 of 32 subjects undergoing biopsy or surgery.
Mean/median size of nodules was 10.0/8.0 mm on T2-STIR, 9.8/8.0 mm on T2, 9.3/8.0 mm on bSSFP, 9.2/7.0 mm on THRIVE, and 9.5/7.5 mm on LDCT. Size measurements significantly correlated among the acquired MRI sequences (p<0.001), and for each sequence the measurements significantly correlated with LDCT (p<0.001). Malignant nodules were significantly larger than benign nodules (average size of malignant nodules 16.38 mm, average size of benign nodules 8.05 mm; p=0.001).
The highest average contrast ratio between malignant and benign nodules was provided by T2-STIR (2.38), followed by T2 (2.29), THRIVE (1.68), and bSSFP (1.42).
Seven of eight subsolid nodules were visible on T2-STIR and T2 with signal intensity and contrast being significantly lower compared to solid nodules (p=0.001 for signal intensity and p=0.002 for contrast on T2-STIR, p<0.001 for signal intensity and contrast on T2). Two of eight subsolid nodules were detected by bSFFP, none by THRIVE. All three calcified nodules were visible on THRIVE, and one was detected by bSSFP; T2-STIR and T2 could not detect any calcified nodules. 4 of 46 nodules (13-35 mm) showed spicules on LDCT; all of them were malignant. The presence of spicules was accurately detected by each MRI sequence.
Consistent with other studies [1, 2, 10], malignant nodules showed higher signal intensity and contrast in our study on T2-weighted images (T2WI), bSSFP, and THRIVE. Calcified nodules were visible on THRIVE but not on T2WI, which matches the statement of two studies that calcification shortens T2 but not T1 relaxation time [15, 16]. Subsolid nodules can turn in carcinomas with lepidic growth. Based on our study results, a subsolid appearance seems to manifest in lower signal intensity on all MRI-sequences. Subsolid nodules were detected with a sensitivity of 88% in our study, which is comparable to Koyama et al. (78%) [10].
T2WI with and without fat suppression were most capable of detecting nodules in our study as well as in different other studies [2, 9, 10]. Based on our results, the T2-STIR sequence provided the highest sensitivity and specificity for nodules in total, and showed the highest contrast for malignant nodules. The advantage of the T2-STIR sequence might be explained by the fact that T1- and T2-contrasts are additive in STIR imaging, which is helpful, since many pathological lesions demonstrate an increase in both T2 and T1 [10, 17]. Even though the T2-STIR sequence was most suitable, it must be considered that MRI sequences are commonly viewed in synopsis, and that a definite diagnosis is normally made, when lesions are visible in more than one sequence.