Magnetic resonance imaging for lung cancer detection: Experience in a population of more than 10,000 healthy individuals

Lung cancer is the leading cause of cancer death worldwide [1–3]. The overall 5-year survival rate is approximately 15% in the United States and less than 10% in Europe [4–6]. Because most lung cancers generate no symptoms at early stages, they are usually diagnosed at an advanced stage and have a poor prognosis. However, many lung cancer deaths could be avoided if tumors were detected at an early stage when they were still resectable. For stage Ia non-small cell lung cancer, the 5-year survival rate is higher than 80%. Prognosis and treatment outcomes have been found to be related to the disease stage at the time of diagnosis [7–9]. Various modalities have been investigated for detecting early lung cancer and consequently reducing lung cancer mortality. In the 1970s, several randomized controlled trials of chest radiography, alone or combined with sputum cytology, were performed. However, because of the poor sensitivity of these methods, the reported data indicated no evidence of benefit in terms of reduction in lung cancer mortality [10–12]. In the 1990s, many observational clinical trials were begun of low-dose spiral computed tomography (LDCT). Those studies showed that although LDCT can detect more early lung cancers than chest radiography, it does not result in a decrease in lung cancer deaths [13–19].

Until recently, magnetic resonance imaging (MRI) was regarded as an inappropriate tool for lung cancer screening because of its insufficient anatomic detail, along with being time-consuming and expensive. The advanced MRI techniques for lungs have reduced the examination time to less than 10 minute (< 8 breath-holds) and improved diagnostic sensitivity and specificity. We conducted a study to assess the feasibility of performing rapid MRI for lung cancer detection. We report our techniques and findings herein.

Of the 49 lung cancers, the mean age of the individual at diagnosis was 60.8 years (range, 36-87 years), and 25 (51.0%) were in males. Eleven cases were in people who had smoked, and the median number of pack-years of smoking was 35 (range, 10-80). Characteristics of individuals with detected primary lung cancers are shown in Table 3. Suspected lung cancers were subjected to various procedures for proper histopathological documentation. Thoracotomy (including video-assisted thoracoscopic surgery) was performed for 40 cancers, lobectomy for 33 cancers, wedge resection for 6, and segmentectomy for 1. Three cancers were histologically proved by bronchoscopic biopsy, 1 by CT-guided lung biopsy, 1 by sputum cytology, 1 by biopsy at supraclavicular lymph node, and 1 by endoscopic biopsy at metastatic duodenal lesion. Two individuals refused any procedures or therapies and later developed rib metastases or progressive pulmonary nodules. Forty cancers had pathological staging, and 9 had clinical staging. Twenty-five cancers were TNM stage Ia disease (51.0%), 12 stage Ib disease (24.5%), 4 stage IIIa (8.2%), 2 stage IIIb (4.1%), and 6 stage IV disease (12.2%). The mean size of detected lung cancers was 1.98 cm (median, 1.5 cm; range, 0.5-8.2 cm). Thirty-six cancers (73.5%) had a primary tumor measuring ≤3 cm, and 29 (59.2%) ≤2 cm. Histological types were as follows: 38 adenocarcinoma (77.6%), 5 bronchioloalveolar carcinoma (10.2%), 2 large cell carcinoma (4.1%), 1 squamous cell carcinoma (2.0%), and 1 carcinoid tumorlet (2.0%). There were 3 female individuals with synchronous double primary lung cancers. One biopsy proved to be non-small cell lung carcinoma with no primary tumor origin in the lungs when retrospectively and prospectively assessed. Two cases did not have tissue proof because the individuals refused any procedures.

Follow-up (duration, 2-9 years) for the 49 cancers showed 8 deaths due to lung cancer, 2 stage IIIa cancers, 2 stage IIIb cancers, and 4 stage IV cancers. One stage Ib individual developed recurrent tumors 2 years later and underwent pulmonary resection again. Thirty-three individuals (67.3%) were alive and free of disease at the time of this writing.

Eight false-positive individuals whose standard-dose CT also suggested malignancies underwent invasive procedures: 2 CT-guided biopsies and 6 thoracotomies (including video-assisted thoracoscopic surgery). Infectious diseases were histologically confirmed in these 8 people, with cryptococcal infections in 6.

To date, several studies have investigated the feasibility of LDCT for screening for primary lung cancers. By enrolling heavy smokers and older people, these studies targeted groups at high risk for lung cancer, resulting in cancer detection rates ranging from 0.4% to 2.7% and lung nodule prevalence rates ranging from 5.1% to 43% [13–19]. In our study, there were no limitations on smoking history and age because all examinations were self-paid and all examinees were self-enrolled. Consequently, our group included a higher proportion of never-smokers and a wider age range (11-94 years). If we calculate the results for smokers between 51 and 70 years old, as did the LDCT studies, the detection rate is 1.4% (6 of 434), and the prevalence of suspicious lung nodules is 9.0% (39 of 434). The proportion of lung cancers among suspicious nodules detected by MRI was 15.4% (6 of 39), higher than in reports for LDCT, which range from 2.2% to 11.6%. While using standard-dose CT as a standard-of reference for nodule detection, previous studies have showed that MRI can achieve high sensitivity for nodules larger than 1 cm, which are clinical approachable [23, 24]. Undoubtedly LDCT is more sensitive than MRI in detecting lung nodules and includes more and smaller nodules. However, for any diagnostic tool, higher sensitivity is usually accompanied by lower specificity. Most of the additional lung nodules detected by LDCT are smaller than 0.5 cm. According to guidelines for managing small lung nodules, nodules smaller than 1 cm are associated with more follow-up visits but not more lung cancers [21, 22]. We believe that it is why our results showed that MRI had a lower prevalence of suspicious nodules than did LDCT but had a cancer detection rate similar to that of LDCT.

Contrasted MRI can detect more morphological details than noncontrasted MRI, and is especially better at depicting peripheral one-third parenchymal vascular markings (around 0.2 cm in diameter). The expected detection limit for contrasted MRI is about 0.3 cm (2 × 2 pixels), better than noncontrasted MRI about 0.5 cm (3 × 4 pixels). Nevertheless, our results showed that no significant difference in cancer detection rates was found between contrasted and noncontrasted MRI (0.5% vs 0.4%, p = 0.24). It was assumed that the proportion of lost to follow-up and false-negative rate of contrasted MRI were similar to that of noncontrasted MRI because all examinations and follow-ups were performed in the same institution. Therefore, the possible explanation might be the same as previous comparison between MRI and LDCT. More subtle lung nodules do not necessarily bring out more lung cancers.

The prevailing argument that MRI is too time-consuming is no longer a valid concern. In this study, we demonstrated how a rapid lung MRI study can be performed with ease. Implementation of the parallel imaging technique has improved the image quality of HASTE in the axial and sagittal views. Currently, we acquire the VIBE sequence in axial, coronal, and sagittal planes by the reduction of TR/TE (2.9 ms/1.1 ms) and a factor of 2 with integrated parallel acquisition techniques (iPAT), which is allowed to cover the entire lung volume within a single breath hold. This optimization not only permits larger coverage but also reduces the pulsation artifact from the heart and large vessels. As shown in Figure 2, VIBE images can display clear branches of pulmonary vessels and bronchial tree with minimal pulsation artifacts. Although the resolution of contrasted MRI is still lower than that of standard-dose CT, MRI can show most nodules around 0.3 cm as standard-dose CT (Figure 3). The entire MRI examination of the lungs can be completed in less than 10 minutes, even if additional scanning is performed for suspicious nodules.

The radiation exposure should be justified when recommending the use of LDCT as a screening tool. Several authors have addressed the radiation hazards of multiple LDCT examinations [25–27]. When lung cancer screening involves the general population and is not targeted just at the high-risk population, the use of MRI can avoid the ethical problems of unnecessary ionizing radiation, which in the long term may contribute to lung cancer risk [28, 29]. In addition, when contrast study is necessary, gadodiamide is noniodinated and has a better safety record with less nephrotoxicity [30].

In our results, there was no significance of cancer detection rate between smokers and never-smokers. However, most studies of lung cancer screening are aimed at smokers only. Many studies have demonstrated that lung cancer in never-smokers is an important public health issue, especially in females who tend to have more non-smoking associated cancers than males [31, 32]. Clearly, to perform lung cancer screening in general population deserves more consideration and further researches.

Although the results of this study provide helpful insight into the use of MRI, the study has several limitations. A certain number of examinees were lost to follow-up. Moreover, follow-up periods and follow-up tools were not identical in all individuals. Standard-dose CT and plain chest radiography were performed only for those with suspected lung lesions; therefore, comparison of MRI with standard-dose CT and chest radiography for all nodules was not possible. Without a reference study, the percentage of false-negative cases could not be evaluated. In addition, like other one-arm observational nonrandomized trials, our study does not provide information on potential stage shift and reduction of lung cancer mortality by screening. Today, several randomized controlled trials of LDCT are under way worldwide [33–35]. We hope that our results will prompt subsequent randomized controlled trials of MRI that are not affected by lead time and length time bias and overdiagnosis.

Based on the results of more than 10,000 MRI examinations, the 1.5T MRI with advanced pulse sequences approach described here, which takes less than 10 minutes to perform, is useful in lung cancer detection for asymptomatic individuals and avoids unnecessary ionizing radiation.