Multiparameter Magnetic Resonance Quantitative Evaluation of Pancreatic Cancer with Vascular Invasion.

Objective: To analyze the value of multiparameter magnetic resonance (mpMRI) in the diagnosis of pancreatic cancer with vascular invasion from two aspects: morphology and function, so as to provide a reliable diagnostic basis for preparing the clinical treatment plans.
Methods: Totally 31 case data of pancreatic cancer patients diagnosed in our hospital from January 2020 to March 2021 were enrolled in this study. All patients underwent multiparameter magnetic resonance imaging (T1WI, T2WI, DKI, and DCE-MRI) before surgery, and then all patients underwent pancreatic cancer surgery. Two experienced radiologists analyzed these obtained images according to the image reports and combined them with the pathological results. Taking pathological results as gold standard, the sensitivity, specificity, and accuracy of quantitative parameters derived from T2WI, DKI, DCE, T2WI + DKI, T2WI + DCE, and T2WI + DKI + DCE for the diagnostic capabilities of pancreatic cancer vascular invasion were calculated using diagnostic laboratory methods. Kappa consistency test was used to estimate the consistency of two radiologists' diagnosis and analysis. The images obtained by DKI sequence were input into the postprocessing software MITK-Diffusion v2014.10.02, The images obtained from DCE sequence were processed by the Tissue 4D software on the Siemens syngo via workstation to calculate and analyze each tumor ROI's MD, MK values from DKI, and K trans, K ep, V e values from DCE. Independent samples t-test was used to compare the parameters of pancreatic cancer with vascular invasion group (16 cases) and nonvascular invasion group (15 cases). ROC curve was used to analyze the efficacy of each parameter in diagnosing pancreatic cancer vascular invasion.
Results: The sensitivity, specificity, and accuracy of T2WI were 62.5%, 53.5%, and 58.1%; those of DKI were 56.3%, 60.0%, and 58.1%; those of DCE were 68.8%, 60.0%, and 64.5%; those of T2WI + DKI were 68.8%, 66.7%, and 67.7%; those of T2WI + DCE were 75.0%, 66.7%, and 71.1%; those of T2WI + DKI + DCE were 81.2%, 73.3%, and 77.4%, respectively. These two diagnostic radiologists analyzed image data with good consistency, Kappa = 0.834. MD, MK, K trans, K ep, and V e were significantly different between the vascular invasion group and the nonvascular invasion group (p < 0.05). Each parameter's AUC of ROC curve was 0.773, 0.829, 0.794, 0.802, and 0.846 (p < 0.05). Take MD = 2.285 × 10-3 mm/s2, MK = 0.72, K trans = 0.103, K ep = 0.337, and V e = 0.353 as thresholds; the sensitivity of these parameters to diagnose vascular invasion of pancreatic cancer was 73.33%, 75%, 87.5%, 68.8%, and 68.8%. The specificity of them was 75%, 80%, 60%, 86.7%, and 86.7%, respectively.
Conclusion: The combined analysis of T2WI + DKI + DCE can improve the specificity and accuracy of diagnostic efficiency of vascular invasion of pancreatic cancer and provide an important diagnostic basis for pancreatic cancer's preoperative treatment.

1. Introduction

Pancreatic cancer has a high mortality rate, and the five-year survival rate does not exceed 5%. According to the location of the lesions, it can be divided into pancreatic head cancer, pancreatic body cancer, and pancreas cancer, of which more than 50% are pancreatic head cancer [1]. Pancreatic cancer has no specific symptoms in the early stage and has a high degree of malignancy. Patients are often diagnosed at an advanced stage and located in the retroperitoneum, making the operation difficult [2]. At present, surgical resection is still the only possible cure for pancreatic cancer, and the 5-year survival rate of early-stage patients after radical surgical resection can reach 25% [3]. Preoperative accurate determination of whether the peripancreatic blood vessels have invaded is the key to determining whether pancreatic tumors can be resected [4]. Currently, magnetic resonance imaging is increasingly being used as the primary test for evaluating pancreatic cancer [5]. If combined with traditional morphological sequences, functional MRI techniques, such as dynamic contrast-enhanced MRI (DCE-MRI) and diffusion-weighted imaging (DWI), additional information on tissue microvascular and cellular architecture can be obtained to better describe tumors [6-9]. Diffusion kurtosis imaging (DKI) is based on non-Gaussian diffusion theory [10] and was initially applied in the neurological field [11]. Because DKI has prominent effectiveness in tumor identification and grading, it has already been widely used in clinical work to diagnose organ diseases such as prostate cancer [12]. At present, studies have explored the application value of multiparameter magnetic resonance imaging (multiparameter MRI, mpMRI) and multislice spiral computed tomography angiography (MSCTA) in the diagnosis of peripancreatic vascular invasion of pancreatic cancer [13, 14], but there are few reports focus on the application of DKI and DCE-MRI in order to assess peripancreatic vascular invasion in pancreatic cancer. The purpose of this study was to explore the value of mpMRI in the diagnosis of peripancreatic vascular invasion in pancreatic cancer and to provide objective imaging evidence for the selection of clinical treatment options.

2. Materials and Methods

2.1. Normal Information

The data of 31 patients were diagnosed as pancreatic cancer with vascular invasion by surgery and pathology in our hospital from January 2020 to March 2021 and successfully completed MRI examinations. MR images including T1WI, T2WI, DKI, and DCE-MRI were retrospectively analyzed. There were 16 cases in the group and 15 cases in the no vascular invasion group. Among them, there were 19 males and 12 females, aged 25-75 years, with an average of 54.6 ± 12.9 years. Inclusion criteria are as follows: (1) no related treatment for pancreatic cancer before MRI examination, (2) good coordination of breath-hold and no motion artifacts in the image, and (3) postoperative pathological biopsy-confirmed pancreatic cancer. This study was approved by the Medical Ethics Committee of the hospital.

2.2. Instruments and Methods

All patients underwent scanning on a 1.5 T superconducting MR system (MAGNETOM Aera, Siemens Healthineers, Erlangen, Germany) with a 32-channel phased-array body coil. Patients were asked to fast for 4 hours before examination. Conventional scanning sequence parameters are as follows: axial T1WI positive and negative phase GRE sequence: TR = 262mms, TE = 2.62 ms, FOV = 400 mm × 320 mm, matrix = 256 × 256, slice thickness = 6 mm, slice interval = 1.2 mm, and NEX = 1; T2WI sequence: TR = 7520 ms, TE = 96 ms, FOV = 400 mm × 320 mm, matrix = 256 × 256, slice thickness = 6 mm, and layer spacing 1.2 mm. Diffusion kurtosis imaging are as follows: TR = 6200 ms; TE = 78 ms; b-value = 0, 250, 500, 750, 100, 1500, and 2000 s/mm2; NEX = 2; FOV = 330 mm × 258 mm; matrix = 192 × 192; slice thickness = 6.5 mm; layer spacing = 1.3 mm; 15 directions were applied for each b-value diffusion sensitive gradient field, and scanning time is 10 min 42 s. Corrected T1-vibe cross-sectional was scanned before contrast agent injection: FOV = 260 mm × 260 mm, slice thickness = 3.6 mm, TR = 5.08 ms, TE = 1.74 ms, average number of times = 8, matrix = 138 × 192, voxel size = 1.9 mm × 1.4 mm × 3.6 mm, and flip angle = 2° and 15°. Dynamic contrast-enhanced MRI parameters are as follows: flip angle = 15°, TR = 4.24 ms, TE = 1.66 ms, matrix = 138 × 192, voxel size = 1.9 mm × 1.4 mm × 3.6 mm, slice thickness = 3.6 mm, and 40 scan periods in total with 7 s/period. Bolus injection of contrast machine Ounaiying after 3 scans, dose = 0.1 mmol/kg, flow rate = 2.5 ml/s, bolus injection immediately after the contrast machine injection 20 ml normal saline flush tube.

2.3. Image Postprocessing and Analysis

2.3.1. mpMRI Image Analysis

Two senior radiologists experienced in abdominal MRI diagnosis have independently read conventional T2WI, DKI, and DCE-MRI images, respectively, without knowing the pathological results. The blood vessels are surrounded by tumor tissue >1/2 diameter; vessels are embedded, occluded, not shown, or seen in filling defects, and the surrounding with collateral formation is regarded as vascular invasion [14].

2.3.2. DKI and DCE-MRI Image Postprocessing

(1)DKI Postprocessing. Save the desired MRI images in DICOM format. The images obtained by DKI sequence were input into the postprocessing software MITK-Diffusion v2014.10.02 (http://mitk.org/wiki/MITK) and calculated to generate mean diffusivity (MD) and mean kurtosis (MK) images representing diffusion information. A radiologist with more than 8 years experience in diagnosing abdominal diseases placed a region of interest (ROI) at the largest layer of the tumor (area 7.12-9.68mm2) and measured the MK and MD values to avoid necrosis, At the cystic degeneration site, all parameters were measured 3 times, and the average value was taken.

(2)DCE Image Postprocessing. Import the T1 data of the flat-sweep small flip angle and the 31 sets of data of dynamic enhancement into Siemens syngo via workstation (syngo Multimodality Workplace). Motion correction and image matching were performed in steps, ROI was selected, and the Tissue 4D analysis software was used to analyze the quantitative data of DCE-MRI, and the pharmacokinetic model was the Tofts two-compartment model. First, combined with T2WI, dynamic enhancement from 31 groups of T1WI was performed. In the image, the scan phase with the most obvious early enhancement and the best contrast with the surrounding tissue was selected as the phase for subsequent analysis, and a radiologist with more than 5 years experience in diagnosing abdominal diseases manually drew the ROI on the dynamic enhanced image, select the largest layer of the tumor and at least 3 or more layers of tissue above and below the tumor for ROI measurement, and try to avoid the surrounding blood vessels. Then, the Ktrans, Kep, and V maps in the ROI tissue were calculated, and the ROI was copied to the corresponding parameter pseudo-color map, and the corresponding Ktrans, Kep, and V values were further measured.

2.4. Statistical Analysis

Statistical analysis was performed using SPSS17.0 software. Taking the pathological results as the gold standard, the sensitivity, specificity, and accuracy of T2WI, DKI, DCE, T2WI + DKI, T2WI + DCE, and T2WI + DKI + DCE in diagnosing pancreatic cancer vascular invasion were calculated by diagnostic experiments. The Kappa consistency test was used to evaluate the consistency of the image data analyzed by the two readers. The obtained images were input into the corresponding software to calculate and analyze the MD, MK, Ktrans, Kep, and V values of each tumor foci. Two independent samples t-test was used to compare the difference of each parameter value between the vascular invasion group and the nonvascular invasion group, and the ROC was drawn. ROC curve was used to analyze the diagnostic efficacy of each parameter for vascular invasion of pancreatic cancer.

3. Results

3.1. MRI Diagnosis of Pancreatic Cancer with Vascular Invasion Results

Two diagnostic imaging physicians analyzed image data in good agreement, Kappa = 0.834.

In all 31 pancreatic cancer cases, there was vascular invasion: T2WI: 25 cases, overestimated 15 cases; DKI: 25 cases, overestimated 16 cases; DCE: 23 cases, overestimated 12 cases; T2WI + DKI: 20 cases, overestimated 9 cases; T2WI + DCE: 19 cases, overestimated 7 cases; T2WI + DKI + DCE: 19 cases, overestimated 6 cases; no vascular invasion: T2WI: 14 cases, underestimated 6 cases; DKI: 14 cases, underestimated 5 cases; DCE: 15 cases, underestimated 6 cases; T2WI + DKI: 15 cases, underestimated 5 cases; T2WI + DCE: 14 cases, underestimated 4 cases; T2WI + DKI + DCE: 14 cases, underestimated 3 cases (Table 1).

Table 1

Comparison of different MRI parameters and their combination in the diagnosis of pancreatic cancer with vascular invasion and pathological results.

Pathological result	Case	T2WI		DKI		DCE		T2WI + DKI		T2WI + DCE		T2WI + DKI + DCE
		+	-	+	-	+	-	+	-	+	-	+	-
+	16	10	6	9	7	11	5	11	5	12	4	13	3
-	15	7	8	6	9	6	9	5	10	5	10	4	11

Remarks: +: with vascular invasion; -: without vascular invasion.

3.2. Efficacy of MRI in the Diagnosis of Pancreatic Cancer with Vascular Invasion

The sensitivity, specificity, and accuracy of T2WI, DKI, DCE, T2WI + DKI, T2WI + DCE, and T2WI + DKI + DCE for the diagnosis of vascular invasion of pancreatic cancer were, respectively, 62.5%, 53.5%, 58.1%, 56.3%, 60.0%, 58.1%, 68.8%, 60.0%, 64.5%, 68.8%, 66.7%, 67.7%, 75.0%, 66.7%, 71.1%, 81.2%, 73.3%, and 77.4% (Table 2). The sensitivity, specificity, and accuracy of T2WI + DKI + DCE in the diagnosis of pancreatic cancer with vascular invasion were higher than those of T2WI, DKI, and DCE alone.

Table 2

Diagnostic efficacy of different MRI parameters and their combination in pancreatic cancer with vascular invasion [% (case), n = 31].

	Sensitivity	Specificity	Positive expected value	Negative expected value	Accuracy
T2WI	62.5 (10/16)	53.5 (8/15)	40.0 (10/25)	57.1 (8/14)	58.1 (18/31)
DKI	56.3 (9/16)	60.0 (9/15)	36.0 (9/25)	64.3 (9/14)	58.1 (18/31)
DCE	68.8 (11/16)	60.0 (9/15)	47.8 (11/23)	60 (9/15)	64.5 (20/31)
T2WI + DKI	68.8 (11/16)	66.7 (10/15)	55.0 (11/20)	66.7 (10/15)	67.7 (21/31)
T2WI + DCE	75.0 (12/16)	66.7 (10/15)	63.2 (12/19)	71.4 (10/14)	71.1 (22/31)
T2WI + DKI + DCE	81.2 (13/16)	73.3 (11/15)	68.4 (13/19)	78.6 (11/14)	77.y (24/31)

3.3. Comparison of DKI Parameters and DCE-MRI Quantitative Parameters between Pancreatic Cancer with Vascular Invasion Group and without Vascular Invasion Group

The comparison results are shown in Table 3. Compared with the group without vascular invasion, the MD value of the group with vascular invasion was lower, and the values of MK, Ktrans, Kep, and V were higher and statistically significant (p < 0.05).

Table 3

Comparison of DKI value and DCE-MRI parameters between groups with or without vascular invasion of pancreatic cancer (x ± s).

Vascular invasion	Number	MD	MK	K trans	K ep	V e
With	16	2.189 ± 0.232	0.753 ± 0.048	0.125 ± 0.022	0.383 ± 0.078	0.377 ± 0.047
Without	15	2.444 ± 0.278	0.698 ± 0.037	0.102 ± 0.022	0.298 ± 0.050	0.307 ± 0.045
F value		0.605	1.175	0	6.496	0.129
p value		<0.05	<0.05	<0.05	<0.05	<0.05

3.4. ROC Curve Analysis of DKI Value and DCE-MRI Quantitative Parameters between Groups with or without Vascular Invasion of Pancreatic Cancer

There were significant differences in MD, MK, Ktrans, Kep, and V between the vascular invasion group and the nonvascular invasion group (p < 0.05); with MD value of 2.285 × 10−3 mm/s2, MK value of 0.72, Ktrans value of 0.103, Kep value of 0.337, and V value of 0.353 as the thresholds, the sensitivities for diagnosing vascular invasion of pancreatic cancer were 73.33%, 75%, and 75%, respectively. The specificity was 75%, 80%, 60%, 86.7%, 86.7%, respectively. See Figure 1.

Figure 1

(a) The ROC curve of MD; (b) the ROC curve of MK; (c) the ROC curve of Ktrans, Kep, and V.

4. Conclusion

Pancreatic cancer is becoming the second leading cause of cancer-related death, with high mortality due to a lack of early detection methods and the inability to successfully treat patients once diagnosed [2]. Therefore, early detection and treatment are extremely important. MRI provides a safe and reliable inspection method for the detection and characterization of pancreatic cancer due to its feasible multisequence imaging, high tissue resolution, and no iodine allergy and ionizing radiation. With the continuous improvement of MRI hardware and software, the research focus of imaging diagnosis of pancreatic cancer with vascular invasion has gradually developed from morphological analysis to molecular and functional MR. It is expected that the pathophysiological changes of the disease can be obtained earlier before the morphological changes, so as to provide imaging evidence for clinical selection of individualized treatment plans, early evaluation of treatment effects, and early detection of tumor metastasis and recurrence.

DCE-MRI is the application of rapid MRI sequence to continuously collect images before, during, and after intravenous injection of contrast agent, showing the information in the process of contrast agent entering the target organ or tissue blood vessel, passing through the capillary bed and finally being removed, and the degree of signal enhancement. It reflects the physical and physiological properties of the target organ or tissue, including tissue perfusion, capillary surface area, and extravascular-extracellular space [15]. Using a suitable pharmacokinetic model, a series of quantitative parameters can be obtained, such as Ktrans, Kep, and V, which can quantitatively analyze the functional state of the microcirculation of tumor tissue. At present, there have been studies on the application of DCE-MRI in the identification of benign and malignant pancreatic masses [16], but there are few reports on the combination of DCE-MRI and DKI to identify the presence or absence of vascular invasion. The quantitative parameters of MRI in pancreatic cancer with peripancreatic vascular invasion provide references for accurate noninvasive diagnosis and treatment. This study found that Ktrans, Kep, and V in the vascular invasion group were higher than those in the normal group. Ktransrepresented the diffusion rate of the contrast agent from the intravascular to the extracellular space outside the blood vessel, and Kep represented the return rate of the contrast agent from the extracellular space into the blood vessel, both of which are related to the permeability of capillaries, andKtransis also affected by the amount of blood perfusion in the tissue and the surface area of capillaries. The occurrence of tumors is accompanied by angiogenesis to meet the requirements of oxygen, nutrient requirements, and metabolic waste removal [17]. Because angiogenesis is too fast and does not have the conditions for normal angiogenesis, many new blood vessels are immature, and immature blood vessels are immature. The increase in perfusion and vascular surface permeability increases. Therefore, pancreatic cancer with poor differentiation and vascular invasion has higher Ktrans and Kep, and V represents the volume percentage of the extracellular space occupied by the extracellular space. The more poorly differentiated tumor, the greater the cell atypia, the tighter the arrangement, and the more immature new blood vessels [18], corresponding to the volume occupied by the extracellular space. The smaller the percentage, the smaller the V value of pancreatic cancer with vascular invasion group, which is inconsistent with the results of this study. It is speculated that it may be related to the small sample size of this study, and the sample size will be further increased in future studies to demonstrate.

The human body is composed of different tissues, and the cell types, cell density, and blood supply of each tissue are different. The diffusion of water molecules is non-Gaussian [10]. The DKI is a single-shot echo plane sequence with seven different b values and three mutually perpendicular scanning directions; the MK value is calculated from the average apparent kurtosis coefficient (AKC) in each direction, the larger the MK value, the more complex the target tissue structure, so it can be used to measure the tissue structure complexity [12]. Malignant lesions are rich in interstitial blood vessels, so it is speculated that the MK value of malignant lesions is higher than that of normal tissues, which is confirmed by the results of this study. The MD value is similar to the ADC value of the DWI single-exponential model, which reflects the diffusion state of water molecules in the human body. The cells of malignant lesions are closely arranged, which restricts the diffusion of water molecules, and the MD value decreases. Therefore, it is speculated that the MD value of malignant lesions should be lower than that of normal tissues; this study has confirmed the above-mentioned inference. The results of this study have also indicated that the DCE-MRI pharmacokinetic parameters and DKI parameters had high efficacy in distinguishing pancreatic cancer with vascular invasion. There were significant differences in MD, MK, Ktrans, Kep, and V between the vascular invasion group and the nonvascular invasion group (p < 0.05). The AUC values of the ROC curve of each parameter were all greater than 0.750, and some even reached more than 0.8. The above results further indicate that DCE-MRI combined with DKI has great potential in early, noninvasive, and overall prediction of whether pancreatic cancer is accompanied by vascular invasion.

There are still some shortcomings in this study: first, the sample size is relatively small. Because pancreatic cancer is insidious and highly malignant, most patients cannot be surgically removed when it is discovered. Therefore, it is relatively difficult to collect pancreatic cancer with surgical and pathological results. To follow-up, we will continue to conduct research in this area and further verify the above results in a large sample; secondly, due to the influence of various factors such as imaging methods, parameters, and image postprocessing, the data obtained by DCE-MRI may vary among institutions, which is also a common problem faced by current DCE-MRI research. Therefore, it is urgent to develop a general DCE-MRI standardized software to meet daily clinical work or organize multicenter research in order to make the obtained study results more reliable and stable.

To sum up, the combined application of T2WI + DKI + DCE could improve the diagnostic performance of pancreatic cancer with vascular invasion and provide a basis for early diagnosis and treatment of pancreatic cancer. MD and MK values derived from DKI and DCE-MRI parameters could quantitatively evaluate pancreatic cancer with vascular invasion. Therefore, in practice, rational use of the morphological features and functional information of multiparameter magnetic resonance imaging can improve the accuracy of pancreatic cancer vascular invasion diagnosis and has a high diagnostic value for pancreatic cancer with vascular invasion.