Assessment of Proton Resonance Frequency Shift Magnetic Resonance Thermography Imaging Quality for Head and Neck Tumors

Introduction

Interstitial thermal ablation can be performed for treating both benign and malignant neoplasm lesions in the head and neck.^1-6 In general, the success of interstitial thermal ablation depends upon accurate insertion of the ablation probe into the target lesion, real-time monitoring of the effects of the treatment, and subsequent evaluation of the extent of thermal tissue damage. ⁶ In order to optimize the safety of the thermal ablation, magnetic resonance (MR) thermography is typically implemented during the procedure.

Magnetic resonance thermography is a validated non-invasive imaging technique that can provide accurate temperature monitoring during thermal ablation.^6,7 Several MR thermography methods have been developed, including sequences based on proton resonance frequency (PRF), the diffusion coefficient, T₁-and T₂-relaxation times, magnetization transfer, and proton density.^7,8 In particular, PRF techniques exploit temperature-induced effects of chemical shift, in which the temperature difference is directly proportional to the phase difference. This method provides accurate temperature measurements in the range of interest for thermal ablation. ⁸

The purpose of this study is to objectively characterize the image quality of phase contrast MR thermography sequences in patients with head and neck tumors for planning interstitial thermal ablation procedures.

Materials and Methods

Patients

This single-institution retrospective study included adult patients with head and neck tumors who underwent magnetic resonance imaging (MRI) exams that included MR thermography sequences. Patients with cutaneous and orbital neoplasms, as well as lipomatous tumors, were excluded from this study. Institutional review board approval was obtained for this study.

MRI Parameters

Axial MR thermography images were acquired on a 3T MRI system (Ingenia dStream; Philips Healthcare, Best, the Netherlands) using a head and neck neurovascular coil in conjunction with routine clinical head and neck MRI sequences. The MR thermography sequence used the following parameters: TR: 30 ms, TE: 10 ms, flip angle: 19.5°, slice thickness: 3.0 mm, number of excitations: 1, matrix size: 252 × 126, and field of view: 240 mm.

Signal-to-Noise Ratio Calculation

Signal-to-noise ratio (SNR) was calculated for each of the head and neck lesions depicted on MR thermography images according to the following equation

$\mathrm{SNR}=\frac{S_L-B{G}_{\mathrm{avg}}}{N_{\text{avg}}}$

(1)

where $S_L$ is the mean of a region of interest (ROI) drawn on the lesion, $B{G}_{\text{avg}}$ is the average of four measurements of the means in background ROIs drawn outside the head region, and $N_{\text{avg}}$ is the average of the measured standard deviations in these same background ROIs. ⁹ Signal-to-noise ratio for all thermography sequences was determined based on measurements obtained at four pixels in the magnitude images, ¹⁰ by dividing the temporal mean by the temporal standard deviation at each pixel.

Contrast-to-Noise Ratio Calculation

For each head and neck lesion, contrast-to-noise ratio (CNR) was calculated from MR thermography images according to the following equation

$\text{CNR}=\left|\frac{S_L-S_{ST}}{N_{\text{avg}}}\right|$

(2)

where $S_L$ is the mean of an ROI drawn on the lesion, $S_{ST}$ is the signal of an ROI in the tissue surrounding of the lesion, and $N_{\text{avg}}$ is the noise average from four background ROIs in the same MR thermography image.

Statistical Analysis

The ranges and standard deviations of the lesion area, SNR, and CNR were determined in each patient. In addition, the Pearson correlation coefficients between SNR and CNR and between lesion area and SNR were calculated. A P-value of less than .05 was considered to be statistically significant.

Results

A total of 12 head and neck lesions in 11 patients with head and neck tumors were assessed, with lesion details and patient demographics listed in Table 1. The average signal-to-noise for the lesions was 482.9 (range: 9.3 to 1973.4; standard deviation: 557.2), while the average CNR for the lesions was 106.7 (range: .03 to 320.6; standard deviation: 99.8) (Table 2). Notably, the highest SNR and CNR was measured in level 2 lymphadenopathy (Figure 1), in contrast to level 2 lymphadenopathy in another patient with relatively low SNR and CNR (Figure 2). The correlation coefficient between SNR and CNR was .6 (P-value: .04) and the correlation coefficient between lesion area and SNR was −.2 (P-value: .6).

Table 1.

Patient demographics and tumors evaluated.

Patient	Lesion site	Histology	Gender	Age
1	Parotid gland	Basaloid salivary gland neoplasm	Male	74
2	Level 6 lymph node	Metastatic esophageal squamous cell carcinoma	Male	53
3	Tongue base and level 2 lymph node	Metastatic oropharyngeal squamous cell carcinoma	Male	30
4	Level 2 lymph node	Metastatic oropharyngeal squamous cell carcinoma	Male	46
5	Level 2 lymph node	Metastatic oropharyngeal squamous cell carcinoma	Male	60
6	Oral tongue	Recurrent oral squamous cell carcinoma	Male	64
7	Level 4 lymph node	Metastatic oral squamous cell carcinoma	Male	46
8	Level 2 lymph node	Metastatic oropharyngeal squamous cell carcinoma	Male	71
9	Parotid gland	Benign cyst	Female	46
10	Retropharyngeal lymph node	Metastatic nasopharyngeal squamous cell carcinoma	Male	53
11	Sinonasal	Sinonasal intestinal-type adenocarcinoma	Male	79

Table 2.

Lesion size, SNR, and CNR.

Patient	Lesion site	Area (mm2)	SNR	CNR
1	Parotid gland	75.6	295.8	152.1
2	Level 6 lymph node	94.73	10.8	.03
3	Tongue base	100.1	1028.6	59.7
	Level 2 lymph node	208.1	1973.4	320.6
4	Level 2 lymph node	125.1	11.6	0.5
5	Level 2 lymph node	40.171	412.5	7.4
6	Oral tongue	479.7	658.3	138.2
7	Level 4 lymph node	338.5	241.9	199.8
8	Level 2 lymph node	292.3	9.3	3.7
9	Parotid gland	186.3	171.3	188.3
10	Retropharyngeal lymph node	59.781	481.4	107.0
11	Sinonasal	125.1	499.9	102.5

Abbreviations: SNR, signal-to-noise ratio; CNR, contrast-to-noise ratio.

Figure 1.

Axial MR thermography image shows left level 2 lymphadenopathy (arrow) with SNR of 1973.4 and CNR of 320.6. The images are sharp and homogeneous. Abbreviations: MR, magnetic resonance; SNR, signal-to-noise ratio; CNR, contrast-to-noise ratio.

Figure 2.

Axial MR thermography image shows right level 6 lymphadenopathy with SNR of 10.8 and CNR of .03. The images are inhomogeneous and there is noticeable pulsation artifact. Abbreviations: MR, magnetic resonance; SNR, signal-to-noise ratio; CNR, contrast-to-noise ratio.

Discussion

This study shows that the image quality of MR thermography of head and neck lesions can vary considerably in terms of SNR and CNR, which are important parameters for performing MR thermography-guided interstitial tumor ablation. Therefore, it is important to obtain this sequence as part of a thermal ablation planning scan to ensure that its quality would be adequate for use during the procedure. For example, lesions with SNR of less than 100 have a temperature uncertainty of greater than 3°C, ¹¹ which was the case for 25% of the lesions in this series.

Interestingly, there was substantial variability among lesions in similar regions, such as the level 2 lymph nodes. On the other hand, good image quality could be achieved in different areas of the head and neck, such as the sinonasal region, despite the possibility of degradation from susceptibility effects at the air-soft tissue-skull base interfaces. In addition, it does not appear that lesion size significantly affects the SNR measurements. However, there was a significant correlation between SNR and CNR of the lesions. Thus, one of these parameters could potentially be used as a surrogate for the other in order to save time and effort for the quality assessment. In particular, SNR could be considered a more relevant parameter because anatomic imaging is otherwise used for the probe insertion and the lesions tend to be more conspicuous on T1-weighted and T2-weighted sequences.

Aside from adipose tissue, the PRF thermal coefficient is relatively unaffected by tissue type and has a minor dependence on thermal history even for coagulated tissue. ¹² Thus, it is unlikely that the variability of SNR and CNR measurements were attributable to the particular lesion histology in this series since fat-containing lesions were excluded. Other factors can affect temperature measurements with PRF, including echo time, tissue susceptibility, electrical conductivity, and external field drift.

A limitation of this study is that not all the tumors in this series would be suitable candidates for thermal ablation in actual practice, regardless of the MR thermography image quality. For example, proximity to critical structures and the size of some of the lesions may dissuade against thermal ablation. Nevertheless, this study provides an overview of potential target lesions at different anatomic sites within the head and neck. Another issue is that only factors might determine overall image quality besides CNR and SNR, such as the possibility of patient motion.

Conclusion

Although phase contrast MR thermography image quality is generally adequate for delineating various tumors in different areas of the head and neck, there can be wide variability in the image quality. Therefore, it is important to include and evaluate the quality of the thermography sequence as part of the MRI performed for thermal ablation planning in order provide reliable temperature mapping.