Combined Metabolo-Volumetric Parameters of 18F-FDG-PET and MRI Can Predict Tumor Cellularity, Ki67 Level and Expression of HIF 1alpha in Head and Neck Squamous Cell Carcinoma: A Pilot Study.

BACKGROUND: Our purpose was to evaluate associations of combined 18F-FDG-PET and MRI parameters with histopathological features in head and neck squamous cell carcinoma (HNSCC).
METHODS: Overall, 22 patients with HNSCC were acquired (10 with G1/2 tumors and 12 with G3 tumors).18F-FDG-PET/CT and MRI was performed and maximum standardized uptake value (SUVmax), total lesion glycolysis (TLG) and metabolic tumor volume (MTV) were estimated. Neck MRI was obtained on a 3 T scanner. Diffusion weighted imaging was performed with estimation of apparent diffusion coefficient (ADC). Perfusion parameters Ktrans,Ve, and Kep were derived from dynamic contrast-enhanced (DCE) imaging. Different combined PET/MRI parameters were calculated as ratios: PET parameters divided by ADC or DCE MRI parameters. The following histopathological features were estimated: Ki 67, EGFR, VEGF, p53, hypoxia-inducible factor (HIF)-1α, and cell count. Spearman's correlation coefficient (p) was used for correlation analysis. P < .05 was taken to indicate statistical significance.
RESULTS: In overall sample, cellularity correlated with SUVmax/ADCmin (P = .558, P = .007), TLG/ADCmin (P = .546, P = .009), and MTV/ADCmin (P = .468, P = .028). MTV/Kep correlated with expression of HIF-1α (P = .450, P = 0,047). In G1/2 tumors, SUVmax/ADCmin correlated with HIF-1α (P = -.648, P = .043); MTV/Kep (P = -.669, P = .034) and TLG/Kep (P = -.644, P = .044) with Ki67. In G3 tumors, cellularity correlated with SUVmax/ADCmin (P = .832, P = .001), SUVmax/ADCmean (P = .741, P = .006), and TLG/ADCmin (P = .678, P = .015). MTV/ADCmin and TLG/ADCmin tended to correlate with HIF-1α.
CONCLUSION: Combined parameters of 18F-FDG-PET and MRI can reflect Ki 67, tumor cellularity and expression of HIF-1α in HNSCC. Associations between parameters of 18F-FDG-PET and MRI and histopathology depend on tumor grading.

Background

Head and neck squamous cell carcinoma (HNSCC) is the most frequent malignancy of the upper aerodigestive tract in humans [1].

Different imaging modalities have been established for diagnosis and monitoring of treatment in HNSCC. Positron emission tomography (PET) with 18F-fluorodeoxyglucose (18F-FDG) is an imaging modality with high sensitivity in the detection of primary tumors and lymph node metastases in HNSCC [2], [3], [4], [5]. Furthermore, 18F-FDG-PET parameters like standardized uptake values (SUV), metabolic tumor volume (MTV), and total lesion glycolysis (TLG) can predict tumor stage and behavior of HNSCC [4], [5], [6]. It has been shown that metabolic tumor activity, measured by 18F- FDG-uptake correlated with T-stage of HNSCC [4]. Also SUV can distinguish well differentiated tumors and poorly differentiated lesions: less well-differentiated tumors showed significantly higher SUVs than better-differentiated tumors [5].

Magnetic resonance imaging, especially diffusion weighted imaging (DWI) and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) are other sensitive imaging modalities in diagnosis of HNSCC. So far, DWI by means of apparent diffusion coefficient (ADC) can predict tumor response to radiochemotherapy [7]. Also ADC values can predict lymphonodal metastasizing in HNSCC [8]. Moreover, ADC can reflect histopathological features of HNSCC, especially proliferation potential and tumor cellularity [9], [10].

DCE-MRI can quantitatively characterize tumor perfusion and is associated with microvessel density in HNSCC [11]. Furthermore, DCE-MRI can predict proliferation activity in HNSCC [11].

Overall, 18F-FDG-PET, DWI and DCE MRI can provide complementary information about biological features like metabolic activity, cellularity, and vascularity in HNSCC [10], [11], [12], [13], [14].

Numerous reports showed that combination of these modalities can better characterize primary tumors and metastatic lesions in HNSCC and estimate tumor behavior [13], [14], [15], [16].

Furthermore, some authors suggested that several 18F-FDG-PET, DWI and DCE MRI parameters can be combined together [17], [18], [19]. For example, Baba et al. calculated a new parameter, namely SUV/ADC, and showed that it had a great potential in differentiation between malignant and benign breast lesions [17]. According to Kim et al., combined parameters of 18F-FDG-PET/MRI could be effective predictors of tumor treatment failure after head and neck cancer surgery [19]. Presumably, combined parameters from 18F-FDG-PET and MRI may increase the diagnostic potential of the imaging and may be better associated with clinically relevant biological parameters in HNSCC than DWI, DCE-MRI and 18F-FDG-PET parameters alone.

The aim of the present study was to evaluate the role of combined18F-FDG-PET/MRI parameters for prediction of different histopathological features in HNSCC.

Methods

This prospective study was approved by the institutional review board (study codes 180–2007, 201–10-12,072,010, and 341–15-05102015).

Patients

For this study, 22 patients, 6 (27%) women and 16 (73%) men, mean age, 55.2 ± 11.0 years, range 24–77 years, with different HNSCC were acquired (Table 1a, Table 1b). Low grade (G1/2) tumors were diagnosed in 10 cases (45%), and high grade (G3) tumor in 12 (55%) patients.

Table 1a

Associations between combined parameters PET/ADC and histopathology in HNSCC

Parameters	EGFR	VEGF	HIF-1α	p53	Ki 67	Cell count
SUVmax/ADCmean	P = .081	P = .411	P = −.2	P = .155	P = .241	P = .403
	P = .729	P = .065	P = .385	P = .504	P = .281	P = .063
MTV/ADCmean	P = .134	P = .009	P = .301	P = −.021	P = −.097	P = .324
	P = .563	P = .971	P = .184	P = .929	P = .668	P = .142
TLG/ADCmean	p = .142	P = .132	P = .226	P = −.018	P = −.015	P = .371
	P = .54	P = .568	P = .325	P = .938	P = .946	P = .089
SUVmax/ADCmin	P = .201	P = .329	P = .056	P = .013	P = .339	p = .558
	P = .382	P = .145	P = .81	P = .955	P = .123	P = .007
MTV/ADCmin	P = .194	P = .022	P = .452	P = −.121	P = .042	p = .468
	P = .401	P = .926	P = .04	P = .602	P = .851	P = .028
TLG/ADCmin	P = .166	P = .193	P = .362	P = −.095	P = .073	p = .546
	P = .471	P = .402	P = .106	P = .683	P = .747	P = .009

Table 1b

Associations between combined parameters PET/DCE and histopathology in HNSCC

Parameters	EGFR	VEGF	HIF-1α	p53	Ki 67	Cell count
SUVmax/Ktrans	P = −.039	P = .12	P = −.296	P = .158	P = .025	P = .032
	P = .87	P = .615	P = .205	P = .506	P = .915	P = .889
SUVmax/Ve	P = .027	P = .235	P = −.25	P = .214	P = .177	P = .171
	P = .91	P = .319	P = .289	P = .366	P = .443	P = .457
SUVmax/Kep	P = −.144	P = .045	P = −.026	P = −.214	P = −.111	P = .056
	P = .544	P = .851	P = .915	P = .366	P = .633	P = .81
MTV/Ktrans	P = .005	P = −.057	P = .056	P = .066	P = −.182	P = .156
	P = .985	P = .811	P = .816	P = .782	P = .429	P = .5
MTV/Ve	P = .104	P = .065	P = .062	P = .05	P = −.09	P = .278
	P = .663	P = .786	P = .796	P = .835	P = .699	P = .223
MTV/Kep	P = −.078	P = −.088	P = .450	P = −.308	P = −.23	P = .091
	P = .743	P = .711	P = .047	P = .186	P = .315	P = .695
TLG/Ktrans	P = .039	P = .066	P = −.002	P = .041	P = −.203	P = .209
	P = .87	P = .781	P = .995	P = .865	P = .377	P = .363
TLG/Ve	P = .126	P = .114	P = .018	P = .062	P = −.052	P = .33
	P = .596	P = .634	P = .94	P = .796	P = .823	P = .144
TLG/Kep	p = −.02	P = −.043	P = .2	P = −.292	P = −.27	P = .064
	P = .935	P = .858	P = .398	P = .212	P = .237	P = .784

Significant correlations are highlighted in bold.

Imaging

18F-FDG-PET/CT

In all 22 patients an 18F-FDG-PET/CT (Siemens Biograph 16, Siemens Medical Solutions, Erlangen, Germany) was performed from the skull to the upper thigh after a fasting period of at least 6 hours. Application of 18F-FDG was performed intravenously with a body weight-adapted dose (4 MBq/kg, range: 168–427 MBq, mean ± std.: 281 ± 62.2 MBq). PET/CT image acquisition started on average 76 minutes (range 60–90 minutes) after 18F-FDG application. Low-dose CT was used for attenuation correction of the PET-data.

On the same day, all 22 patients also underwent a whole body simultaneous 18F-FDG PET/MRI (Biograph mMR - Biograph, Siemens Health Care Sector, Erlangen, Germany). Since simultaneous18F-FDG PET/MRI was the secondary imaging modality in the majority of the cases, start of PET/MRI image acquisition time was very inhomogeneous and varies up to 300 minutes post-injection. Since SUV values may be slightly influenced by the time-delay between the PET/CT and PET/MRI investigation due to radiotracer-clearance or further uptake, we decided to not include 18F-FDG-PET-data of simultaneous PET/MRI for the current analysis to guarantee a homogenous group of patients with 18F-FDG-PET image acquisition starting in a range of 60–90 minutes post-injection.

PET/CT image analysis was performed on the dedicated workstation of Hermes Medical Solutions, Sweden. For each tumor, maximum and mean SUV (SUVmax; SUVmean), TLG and MTV were determined on PET-images. Prior to this, tumor margins of the HNSCC were identified on diagnostic CT and MRI and fused PET/CT images and a polygonal volume of interest (VOI), that include the entire lesion in the axial, sagittal and coronal planes, was placed in the PET dataset (SUVmax threshold 40%) (Figure 1A-D).

Figure 1

Imaging findings and histopathological features in a patient with metastatic HNSCC of the left oropharynx. Lesion with polygonal volume of interest (VOI, red area) in the axial (A), coronal (B) and sagittal (C) 18F-FDG-PET planes. SUVmax = 22.6, metabolic tumor volume (MTV) = 18.12, and total lesion glycolysis (TLG) = 255.2.

D. Fused 18F-FDG-PET/CT image of the lesion. E. ADC map of the tumor. The ADC values (× 10–3 mm2 s−1) of the lesion are as follows: ADCmin = 0.98 and ADCmean = 1.5. F-H. DCE MRI images of the tumor: Ktrans = 0.35 min−1 (f), Ve = 0.7% (g), KeP = .61 min−1 (h).

Histopathological parameters are as follows: I. MIB-1 staining. KI 67 index is 45%. Cell count is 121. J. EGFR staining. Stained area is 99,841 μm2. K. VEGF staining. Stained area is 373 μm2. L. HIF-1α staining. Stained area is 14,896 μm2. M. p53 staining. Stained area is 0 μm2.

Diffusion-Weighted Imaging

In all patients, neck MRI was performed on a 3 T MR scanner using a combined head and neck coil. Besides anatomical sequences, an axial DWI EPI (echo planar imaging) sequence with b-values of 0 and 800 s/mm2 (TR/TE: 8620/73 ms, slice thickness: 4 mm, and voxel size: 3.2 x 2.6 x 4.0 mm) was performed. ADC maps were automatically generated by the implemented software (Figure 1C). Regions of interest (ROI) were manually drawn on the ADC maps along the contours of the tumor on each slice (whole tumor measure, Figure 1E). In all lesions minimal ADC values (ADCmin) and mean ADC values (ADCmean) were estimated [10].

Dynamic Contrast-Enhanced Imaging

Dynamic contrast-enhanced (DCE) imaging was performed using T1w DCE sequence (TR/TE 2.47/0.97 ms, slice thickness 5 mm, flip angle 8°, voxel size 1.2 × 1.0 × 5.0 mm) according to our previous description [11], [12]. T1w DCE included 40 subsequent scans à 6 seconds. After the fifth scan, contrast medium (0.1 mmol Gadobutrol per kg of bodyweight (Gadovist, Bayer Healthcare, Leverkusen, Germany)) was administrated of started at a rate of 3 ml per second. Thereafter, the acquired images were transferred to a software module for tissue perfusion estimation (Tissue 4D, Siemens Medical Systems, Erlangen, Germany) as reported previously [11], [12]. The following pharmacokinetic parameters were calculated (for exemplary parameter images see Figure 1F-H) [11], [12]:

Ktrans or volume transfer constant representing vessel permeability. This parameter estimates the diffusion of contrast medium from the plasma through the vessel wall into the interstitial space.

Ve or volume of the extravascular extracellular leakage space (EES);

Kep or parameter for diffusion of contrast medium from extravascular extracellular leakage space back to the plasma.

Combined parameters

In every case, the following combined PET/ADC parameters were calculated according to previous descriptions [17], [19]:

SUVmax divided by ADCmin (SUVmax/ADCmin),

SUVmax divided by ADCmen (SUVmax/ADCmean),

TLG divided by ADCmin (TLG/ADCmin),

TLG divided by ADCmean (TLG/ADCmin),

MTV divided by ADCmin (SUVmax/ADCmin),

MTV divided by ADCmean (SUVmax/ADCmean).

Furthermore, also different combined parameters PET/DCE MRI were calculated [18]. There were parameters between PET and Ktrans:

SUVmax divided by Ktrans (SUVmax/Ktrans),

TLG divided by Ktrans (TLG/Ktrans),

MTV divided by Ktrans (MTV/Ktrans).

Additionally, combined parameters based on associations between PET findings and Ve were calculated:

SUVmax divided by Ve (SUVmax/Ve),

TLG divided by Ve (TLG/Ve),

MTV divided by Ve (MTV/Ve).

Finally, calculation of combined parameters based on associations between PET findings and Kep was also made:

SUVmax divided by Kep (SUVmax/Kep),

TLG divided by Kep (TLG/Kep),

MTV divided by Kep (MTV/Kep).

Histopathological Findings

In all cases, the diagnosis was confirmed histopathologically by tumor biopsy. The biopsy specimens were deparaffinized, rehydrated and cut into 5 μm slices.

The following histopathological features of the tumors were estimated (1i-m):

expression of Ki 67;

expression of epidermal growth factor receptor (EGFR);

expression of vascular endothelial growth factor (VEGF);

expression of tumor suppressor gene protein p53;

expression of hypoxia-inducible factor (HIF)-1α;

cell count.

All stained specimens were digitalized by using the Pannoramic microscope scanner (Pannoramic SCAN, 3DHISTECH Ltd., Budapest, Hungary) with Carl Zeiss objectives up to 41x bright field magnification by default. In the used bottom-up approach, the whole sample was acquired at high resolution. Via Pannoramic Viewer 1.15.4 (open source software, 3D HISTECH Ltd., Budapest, Hungary) the slides were evaluated and three captures with a magnification of x200 were extracted of each sample as reported previously [20].

Further analyses of the digitalized histopathological images were performed by using the ImageJ software 1.48v (National Institutes of Health Image program) with a Windows operating system [10], [11].

Statistical Analysis

Statistical analysis was performed using SPSS package (IBM SPSS Statistics for Windows, version 22.0, Armonk, NY: IBM corporation). Collected data were evaluated by means of descriptive statistics.

Spearman's correlation coefficient (p) was used to analyze associations between investigated parameters. P < .05 was taken to indicate statistical significance.

Results

In overall sample, cell count correlated statistically significant with SUVmax/ADCmin (P = .558, P = 0,007), TLG/ADCmin (P = .546, P = 0,009), and MTV/ADCmin (P = .468, P = 0,028) (Table 1a). Furthermore, MTV/Kep correlated with expression of HIF-1α (P = .450, P = .047) (Table 1b). There were no statistically significant correlations between other parameters.

In G1/2 tumors, SUVmax/ADCmin correlated well with expression of HIF-1α (P = −.648, P = .043) (Table 2a). Furthermore, MTV/Kep (P = −.669, P = .034) and TLG/Kep (P = −.644, P = .044) correlated with expression of Ki 67 (Table 2b). None of the combined parameters showed statistically significant correlations with cell count.

Table 2a

Associations between combined parameters PET/ADC and histopathology in grade 1/2 tumors

Parameters	EGFR	VEGF	HIF-1α	p53	Ki 67	Cell count
SUVmax/ADCmean	P = −.176	P = .515	P = −.794	P = .248	P = .055	P = −.103
	P = .627	P = .128	P = 0,006	P = .489	P = .88	P = .777
MTV/ADCmean	p = .091	P = −.127	P = .115	P = −.418	P = −.62	P = .491
	P = .803	P = .726	P = .751	P = .229	P = .056	P = .15
TLG/ADCmean	P = −.042	P = −.055	P = −.018	P = −.333	P = −.497	P = .527
	P = .907	P = .881	P = .96	P = .347	P = .144	P = .117
SUVmax/ADCmin	P = −.152	P = .467	P = −.648	P = .067	P = .006	P = .139
	P = .676	P = .174	P = .043	P = .855	P = .987	P = .701
MTV/ADCmin	P = .152	P = −.055	P = .164	P = −.406	P = −.485	P = .624
	P = .676	P = .881	P = .651	P = .244	P = .156	P = .054
TLG/ADCmin	P = .079	P = −.018	P = −.079	P = −.285	P = −.411	P = .527
	P = 0,829	P = .96	P = .829	P = .425	P = .238	P = .117

Significant correlations are highlighted in bold.

Table 2b

Associations between combined parameters PET/DCE and histopathology in grade 1/2 tumors

Parameters	EGFR	VEGF	HIF-1α	p53	Ki 67	Cell count
SUVmax/Ktrans	p = −.176	P = .261	P = −.467	P = .236	P = −.166	P = −.273
	P = .627	P = .467	P = .174	P = .511	P = .647	P = .446
SUVmax/Ve	p = −.176	P = .176	P = −.43	P = .358	P = 0,215	p = −0,248
	P = .627	P = .627	P = .214	P = .31	P = .551	P = .489
SUVmax/Kep	P = −.006	P = .152	p = −.467	P = −.103	P = −.509	P = −.006
	P = .987	P = .676	P = .174	P = .777	P = .133	P = .987
MTV/Ktrans	P = −.03	P = −.164	P = .103	P = −.297	P = −.583	P = .406
	P = .934	P = .651	P = .777	P = .405	P = .077	P = .244
MTV/Ve	P = .115	P = −.115	P = .115	P = −.224	P = −.35	P = .418
	P = .751	P = .751	P = .751	P = .533	P = .322	P = .229
MTV/Kep	p = −.03	P = −.079	p = .079	p = −.418	P = −.669	P = .527
	P = .934	P = .829	P = .829	P = .229	P = .034	P = .117
TLG/Ktrans	p = .042	p = −.018	p = −.03	P = −.188	P = −.497	P = .321
	P = .907	P = .96	P = .934	P = .603	P = .144	P = .365
TLG/Ve	P = .055	p = .055	P = −.139	P = −.091	P = −.215	P = .345
	P = .881	P = .881	P = .701	P = .803	P = .551	P = .328
TLG/Kep	p = .164	p = −.042	p = −.127	P = −.382	P = −.644	P = .418
	P = .651	P = .907	P = .726	P = .276	P = .044	P = .229

Significant correlations are highlighted in bold.

In G3 tumors, cell count correlated statistically significant with SUVmax/ADCmin (P = .832, P = .001), SUVmax/ADCmean (P = .741, P = .006), and TLG/ADCmin (P = .678, P = .015) (Table 3a). Additionally, MTV/ADCmin and TLG/ADCmin tended to correlate with expression of HIF-1α (for each parameter, P = .6, P = .051). None of the PET/DCE parameters had significant correlations with the investigated histopathological features (Table 3b). Only MTV/Kep tended to correlate with expression of HIF-1α (P = .612, P = .06).

Table 3a

Associations between combined parameters PET/ADC and histopathology in grade 3 tumors

Parameters	EGFR	VEGF	HIF-1α	p53	Ki 67	Cell count
SUVmax/ADCmean	P = .364	P = .381	P = −.082	p = .1	P = .31	P = .741
	P = .272	P = .247	P = .811	P = .77	P = .327	P = .006
MTV/ADCmean	P = .318	P = .153	P = .282	P = .518	P = .077	P = .175
	P = .34	P = .654	P = .401	P = .102	P = .811	P = .587
TLG/ADCmean	p = .418	P = .343	P = .245	P = .327	P = .113	P = .441
	P = .201	P = .301	P = .467	P = .326	P = .727	P = .152
SUVmax/ADCmin	p = .327	p = .21	P = .327	P = −.073	p = .5	p = .832
	P = .326	P = .536	P = .326	P = .832	P = .098	P = .001
MTV/ADCmin	P = .255	P = .105	p = .6	P = .273	P = .324	P = .364
	P = .45	P = .759	P = .051	P = .417	P = .304	P = .245
TLG/ADCmin	p = .318	P = .315	P = 0.6	P = .145	p = .31	P = .678
	P = .34	P = .346	P = .051	P = .67	P = .327	P = .015

Significant correlations are highlighted in bold.

Table 3b

Associations between combined parameters PET/DCE and histopathology in grade 3 tumors

Parameters	EGFR	VEGF	HIF-1α	p53	Ki 67	Cell count
SUVmax/Ktrans	p = .042	p = −.006	P = −.503	P = .188	P = −.073	p = .091
	P = .907	P = .986	P = .138	P = .603	P = .83	P = .79
SUVmax/Ve	p = .321	P = .175	P = −.309	p = .212	P = .037	P = .427
	P = .365	P = .63	P = .385	P = .556	P = .915	P = .19
SUVmax/Kep	p = −.152	P = −.123	P = .188	P = −.261	P = .119	P = .118
	P = .676	P = .735	P = .603	P = .467	P = .727	P = .729
MTV/Ktrans	p = .079	P = −.032	P = −.2	P = .503	P = −.119	p = −.164
	P = .829	P = .929	P = .58	P = .138	P = .727	P = .631
MTV/Ve	p = .164	P = .084	p = −.091	P = .418	P = −.083	P = −.036
	P = .651	P = .817	P = .803	P = .229	P = .809	P = .915
MTV/Kep	p = −.055	P = −.084	p = .612	p = .042	p = .037	P = −.318
	P = .881	P = .817	P = .06	P = .907	P = .915	P = .34
TLG/Ktrans	P = .224	P = .071	p = −.103	P = .394	P = −.046	p = .091
	P = .533	P = .845	P = .777	P = .26	P = .893	P = .79
TLG/Ve	p = .212	p = .123	p = −.055	P = .321	P = −.064	p = .164
	P = .556	P = .735	P = .881	P = .365	P = .851	P = .631
TLG/Kep	P = .006	P = −.032	P = .43	p = −.055	P = −.028	P = −.227
	P = .987	P = .929	P = .214	P = .881	P = .936	P = .502

Discussion

Our study showed that combined PET/MRI parameters can reflect different histopathological findings in HNSCC and, therefore, can be used as surrogate markers for tumor characterization.

Previously, some studies investigated associations between several imaging findings and histopathology in HNSCC. However, the reported results were inconclusive. While some authors observed significant associations between imaging and histological parameters in HNSCC, others did not [10], [11], [21], [22], [23], [24]. Recently, a meta-analysis regarding correlations between different imaging parameters and histopathological features in HNSCC was published [25]. It showed that SUV derived from 18F-FDG PET did not correlate with Ki 67(the pooled correlation coefficient was 0.20) [25]. Furthermore, no correlation was observed between SUV and expression of p53 (pooled correlation coefficient = 0.0). However, SUV correlated moderately with expression of HIF-1α (pooled correlation coefficient = 0.44) [25]. Regarding other imaging parameters, a statistically significant correlation between Ktrans and Ki 67 was calculated (pooled correlation coefficient = −0.68) [25]. Also, ADC correlated well with Ki 67 (pooled correlation coefficient = −0.61) [25].

As mentioned above, some authors analyzed combined parameters from 18F-FDG PET and DWI and as well from 18F-FDG PET and DCE-MRI [17], [18], [19]. Overall, there were two studies about combined parameters in breast cancer [17], [18] and one in HNSCC [19]. It has been shown that triple negative breast cancers showed higher metabolic–perfusion ratios, namely SUVmax/Ktrans, MTV/Ktrans, TLG/Ktrans, and TLG/Ve, compared to non-triple negative breast cancers [18]. Furthermore, Baba et al. found that the combination of SUV and ADC, namely SUV/ADC, was more accurate than either parameter alone for differentiating benign from malignant breast lesions [17]. In HNSCC, MTV/ADCmean and TLG/ADCmean can predict tumor recurrence after surgical therapy, as it was shown in a study of Kim et al. [19]. Moreover, Kim et al. also showed that TLG/ADCmean can predict disease-free interval and that TLG/ADCmean and MTV/ADCmean were associated with lymphatic invasion in HNSCC [19].

We assumed that combined parameters of 18F-FDG-PET and MRI may be more sensitive than each parameter alone in reflection of histopathological features in HNSCC. In fact, our results confirmed this hypothesis. As shown, three combined PET/MRI parameters SUVmax/ADCmin, TLG/ADCmin, and MTV/ADCmin correlated statistically significant with tumor cellularity. Furthermore, MTV/Kep correlated with expression of HIF-1α.

The observed associations between the combined parameters are stronger than those reported previously for PET and/or MRI parameters. For example, it has been shown that SUVmax did not correlate significantly with expression of Ki 67, VEGF, EGFR, HIF-1α, and p53 in HNSCC [26]. Furthermore, also the calculated PET/MRI parameters correlated better with cell count and expression of Ki 67 than ADC and/or DCE MRI parameters [27], [28].

We observed also another interesting finding, namely different associations between the analyzed combined PET/MRI parameters and histopathology in dependence on tumor grading. In G1/2 tumors, SUVmax/ADCmin reflected expression of HIF-1α and MTV/Kep and TLG/Kep correlated strongly with expression of KI 67. In G3 tumors, SUVmax/ADCmin, SUVmax/ADCmean, and TLG/ADCmin correlated strongly with tumor cellularity. It is still unknown, why associations between imaging parameters and histopathology depended on tumor grading. Presumably, tumor architecture like ratio parenchyma/stroma is different in well, moderately and poorly differentiated tumors that results in different associations between imaging and histopathology. Previously, some reports observed similar findings. For instance, it has been shown that tumor grading influenced relationships between metabolic activity, perfusion and diffusion in HNSCC [12]. Furthermore, in meningiomas, ADC correlated stronger with cellularity in grade 2/3 tumors than in grade 1 lesions [29].

Our findings have high clinical relevance. Proliferation index Ki67 is an established biomarker in HNSCC to predict tumor behavior and prognosis. High expression of Ki 67 correlated with tumoral aggressiveness and worse prognosis in patients with HNSCC [30], [31]. Therefore, the possibility to estimate proliferation activity based on imaging findings is very important. Also prediction of tumor cellularity on imaging is clinically relevant and may be helpful to evaluate therapy response. Furthermore, imaging parameters may predict expression of HIF-1α, which characterizes cellular responses to hypoxic stress [32]. According to the literature, overexpression of HIF-1α is associated with increase of mortality risk and worse prognosis of HNSCC [32]. Our data suggest that the analyzed combined parameters may be used as surrogate markers in HNSCC. However, this does not apply for all histopathological biomarkers. Neither in the overall sample, nor in the subgroups the combined PET/MRI parameters correlated statistically significant with expression of p53, VEGF and EGFR. EGFR regulates many cellular processes like including proliferation, apoptosis, and differentiation [33]. It has been shown that EGFR expression can represent a good prognostic parameter in HNSCC [33], [34]. Another histopathological parameter, namely tumor suppressor protein p53 regulates the activity of pathways, which lead variously to cell cycle arrest, senescence, or apoptosis following exposure of cells to endogenous or exogenous cellular stresses [35]. Finally, VEGF plays also a great role in HNSCC. It mediated different processes like endothelial cell proliferation, tumoral invasion, cell migration, chemotaxis of bone marrow derived progenitor cells, vasodilation and vascular permeability [36]. As reported previously, overexpression of this marker is a poor predictor for patients with HNSCC [30], [36]. Theoretically, imaging may be associated with the mentioned biomarkers. However, as seen, combined PET/MRI parameters cannot be used for prediction of expression of p53, VEGF and EGFR. Our results are in agreement with some previous reports, in which also no associations between these histopathological biomarkers and different imaging parameters were identified [21], [22], [23], [24], [25].

The results of the present study are limited to a relatively small number of the acquired patients. However, this is the first report regarding associations between combined PET/MRI parameters and complex histopathological features in HNSCC. Clearly, further studies with more patients are needed to confirm our finding.

In conclusion, combined PET/MRI parameters can reflect different histopathological features, in particular KI 67, tumor cellularity and expression of HIF-1α, in HNSCC, and, therefore, can be used as surrogate biomarkers.