PET-MRI provides incremental value in patients with head and neck cancer.

Objectives: There is growing interest in the use of integrated PET-MRI for management of patients with head and neck cancer (HNC). Our objective was to compare the sensitivity and specificity of PET-CT (Siemens, Biograph mCT) and PET-MRI (Siemens, Biograph mMR) using 18F-Fluorodeoxyglucose (FDG) at initial staging and post-therapy evaluation for relapse in HNC patients.

Methods: This retrospective study included 40 HNC patients, 17 at initial diagnosis and 23 at follow-up after treatment. All patients underwent sequential whole body PET-CT and integrated PET-MRI of the head and neck with a 30 min interval after a single injection of FDG, followed by a dedicated head and neck MRI (31 with and 9 without gadolinium contrast) on the same day. All imaging modalities were analyzed separately and blindly. A 5-point scale was used for each modality to classify regional lesions (RL) and lymph nodes (LN) into likely benign, probably benign, indeterminate, probably malignant and likely malignant. The first 2 and the last 3 categories were combined as negative and positive groups for malignancy, respectively. Image artifacts and the lesion SUVmax were comparatively investigated on PET-CT and PET-MRI. Results: Forty HNC patients (24 squamous cell carcinoma and 16 other HNC) had total of 117 abnormal sites; 49 RLs and 68 LNs. Based on histopathology or 3-6 month follow-up imaging, final diagnosis of malignant versus benign etiology was available for 104 sites. The sensitivity and specificity of distinguishing malignant from benign lesions were 89% and 38% for PET-CT, 76% and 85% for MRI and 88% and 87% for PET-MRI, respectively (tables 1-3). There were 17 indeterminate findings on PET-CT (4 RLs and 13 LNs); in 16/17 PET-MRI was helpful to specify the abnormality. There were 21 false positive results on PET-CT which were true negative on PET-MRI (7 RLs and 14 LNS), all confirmed by pathology results or follow-up imaging. PET-MRI showed significantly less hardware artifacts and corrected PET-CT artifacts in 32.5% (P<0.001) of all cases, 39% of follow-up cases (P<0.001) and 23.5% of initial staging cases (P=0.03). On the other hand, PET-CT because of being a whole body imaging modality detected pulmonary metastases in 2 patients, second primary lung malignancy in 1 patient, inflammatory lung nodules in 12 patients, osseous metastases in 1 patient and colonic neoplasia in 1 patient. There was a good SUVmax correlation between PET-CT and PET-MRI but systematically higher by PET-CT. The mean difference was 1.71 for all cases (95% CI 1.09-2.33, p<0.001), 1.94 for initial staging (95% CI 0.80-3.08, p=0.001) and 1.50 for follow-up cases (95% CI 0.90-2.10, p<0.001). The optimal SUVmax cutoff for PET-CT to diagnose malignancy was 6.7 (sensitivity 68%, specificity 80, AUC 0.82) and for PET-MRI, corresponding value was 5.1 (sensitivity 67%, specificity 82, AUC 0.83). Conclusion: The results of this study suggest that integrated PET-MRI provides incremental value and is potentially superior to PET-CT and MRI at initial staging. It also corrects false positive findings on PET/CT by defining unclear FDG uptake during follow-up. However, PET-CT has the advantage of detecting metastatic/second primary malignant lesions in the rest of body.