Performance of Lung MRI Combined to Synthetic CT in the Follow-up of Lung Nodules

Lung cancer is the leading cause of cancer-related deaths worldwide. In France, its incidence was estimated at 46,300 in 2018. In most cases, the diagnosis is initially made by the detection of a nodule or mass on chest X-ray or CT scan. Thus, most often non-invasive follow-up by chest CT scans is recommended. More expensive and invasive methods may also be proposed. However, patients with benign nodules may undergo diagnostic methods that are not without risks (exposure to ionizing radiation, complications related to trans-thoracic or surgical biopsy, etc.).

Lung Cancer Screening Trials (NLST, NELSON) have shown that lung cancer related mortality is reduced in subjects with high risk of lung cancer screened by using low-dose chest CT. Nevertheless, published systematic reviews and meta-analyses report a number of side effects of screening related to false positives and over diagnosis. In addition, the assessment of the risks related to the cumulative dose of exposure to ionising radiation during successive rounds of screening remains unknown. Consequently, the French National Authority for Health (HAS) recommends that pilot programs to be conducted to evaluate the different modalities for the organization of a national lung cancer screening program.

The spatial resolution of magnetic resonance imaging (MRI) of the lung has been significantly improved in the last decade, thanks to the development of ultra-short echo time (UTE) sequences. The advantage of MRI, in addition of being a free-radiation imaging technique, lies in its multiparametric nature with T1-weighted, T2-weighted and diffusion-weighted imaging providing images of different contrasts allowing the characterization of lesions. However, the follow-up of lung nodules, especially with the calculation of the volume doubling time (VDT) on UTE MRI, has not been evaluated. In addition, the performance of multiparametric MRI combining T2 signal, apparent diffusion coefficient (ADC) and nodule volume in determining nodule malignancy remains to be assessed. Recently, the development of artificial intelligence (AI) techniques with generative adversarial networks (GANs) has made it possible to generate CT-like imaging from MRI images. A very recent work demonstrated that AI model is able to generate from UTE lung MRI images, a high resolution synthetic CT image with a very similar texture to the standard CT and better quality than UTE alone. Therefore, the present sudy hypothesis is that multiparametric MRI combined with synthetic CT could have a complementary role with low-dose CT in lung cancer screening to reduce the false positive rate and to perform a free-radiation follow-up of lung nodules

• Area under the ROC curve [Срок оценки: Baseline, Month 3, Month 12]
• Intraclass correlation coefficient between CT and synthetic CT [Срок оценки: Baseline, Month 3, Month 12]
• multiparametric MRI characteristics [Срок оценки: Baseline, Month 3, Month 12]