From the 1980s to the early 1990s, the potentials of using magnetic resonance (MR) imaging to evaluate different lung diseases as well as mediastinal, pleural and cardiac diseases were tested by many physicists and radiologists. At that time, it was concluded that MR imaging could not be used as substitute for computed tomography (CT) because MR systems, sequences and other applications at that time were very primitive and limited for obtaining adequate image quality within an appropriate examination time. Since that time, chest MR imaging is still one of the challenging fields from the academic and clinical points of view. However, recent technical advances including MR system and sequence as well as post-processing software, state of the art pulmonary MR imaging can provide not only morphological, but also functional information in some cardiopulmonary diseases. In addition, MRI with ultrashort TE (UTE) has been suggested as having a potential to demonstrate lung structures and morphological assessment as well as CT in last a few years (1, 2). However, no report has been found to compare the capability of thin-section UTE MRI for pulmonary nodule detection and nodule type assessment to that of thin-section low- and standard-dose CTs.

We hypothesized that newly developed pulmonary thin-section MRI with UTE has a similar potential to detect pulmonary nodules and evaluate nodule types as well as thin-section low- and standard-dose CTs. The purpose of this study was to compare the capability of pulmonary MRI with UTE for detection of lung nodules and evaluation of nodule type with thin-section low- and standard-dose CTs.

170 consecutive patients (96 males: mean age, 70 years and 74 females: mean age, 70 years) with suspected pulmonary nodules at near-by hospitals were examined with chest standard- and low-dose CTs and pulmonary MR imaging with UTE. In 170 patients, total 274 nodules (195 solid nodules and 79 sub-solid nodules <63 ground-glass nodules and 16 part-solid nodules>) were detected on standard-dose CT by consensus of board-certified chest radiologists. All low- and standard-dose thin-section CTs (LDCT and SDCT) were performed by a 320-detector row CT (Aquilion ONE, Toshiba Medical Systems Corporation, Otawara, Tochigi, Japan), and all pulmonary MRI with UTEs were examined by a 3T MR system (Vantage Titan 3T, Toshiba) by respiratory-gated 3D radial UTE pulse sequence (TR 4.0 ms/ TE 110-192 μs, flip angle 5 degree, 1x1x1 mm3 voxel size). Then two chest radiologists with more than 10 year experiences evaluated capabilities for nodule detection and subtype classification by visual assessment. The probability of nodule presence on each method was assessed by means of 5-point scale on a per nodule basis. In addition, the nodule subtype on each method was also evaluated by means of 4-point scale on a per nodule basis.

To compare the capability for nodule detection among all methods, weighted jackknife free-response receiver operating curve (JAFROC) was performed. Then, sensitivity was compared among all methods by means of McNemar’s test. To compare inter-method agreement for nodule subtype classification among all methods, κ statistics and χ2 tests were performed. A p value less than 0.05 was considered as significant in this study.

Representative case is shown in Figure 1, 2 and 3.

Results of JAFROC analysis are shown in Figure 4. Figure of merit (FOM) of all methods (SDCT: FOM=0.96, LDCT: FOM=0.95, MRI with UTE: FOM=0.97) had no significant differences (F=0.46, p=0.67). In addition, sensitivities of all methods (SDCT: 96.9%, LDCT: 95.3%, MRI with UTE: 94.5%) had no significant differences (p>0.05).

When assessed inter-method agreements among all methods, inter-method agreements were also almost perfect (SDCT vs. LDCT: κ=0.98, p<0.0001; SDCT vs. MRI with UTE: κ=0.82, p<0.0001; LDCT vs. MRI with UTE: κ=0.84, p<0.0001).