Could new reconstruction CT techniques challenge MRI for the detection of brain metastases in the context of initial lung cancer staging?

Objectives To evaluate the diagnostic performance of brain CT images reconstructed with a model-based iterative algorithm performed at usual and reduced dose. Methods 115 patients with histologically proven lung cancer were prospectively included over 15 months. Patients underwent two CT acquisitions at the initial staging, performed on a 256-slice MDCT, at standard (CTDIvol: 41.4 mGy) and half dose (CTDIvol: 20.7 mGy). Both image datasets were reconstructed with filtered back projection (FBP) and iterative model-based reconstruction (IMR) algorithms. Brain MRI was considered as the reference. Two blinded independent readers analysed the images. Results Ninety-three patients underwent all examinations. At the standard dose, eight patients presented 17 and 15 lesions on IMR and FBP CT images, respectively. At half-dose, seven patients presented 15 and 13 lesions on IMR and FBP CT images, respectively. The test could not highlight any significant difference between the standard dose IMR and the half-dose FBP techniques ( p -value = 0.12). MRI showed 46 metastases on 11 patients. Specificity, negative and positive predictive values were calculated (98.9–100 %, 93.6–94.6 %, 75–100 %, respectively, for all CT techniques). Conclusion No significant difference could be demonstrated between the two CT reconstruction techniques. Key points • No significant difference between IMR100 and FBP50 was shown. • Compared to FBP, IMR increased the image quality without diagnostic impairment. • A 50 % dose reduction combined with IMR reconstructions could be achieved. • Brain MRI remains the best tool in lung cancer staging.