Diffusion weighted imaging (DWI), with quantitative measurement of apparent diffusion coefficient (ADC) values, is routinely performed in clinical practice in detection and characterization of pancreatic diseases [1]. Differentiating mass-forming focal pancreatitis (FP) and pancreatic ductal adenocarcinoma (PDAC) is of great clinical importance as distinctive treatments are usually carried out. For now, there is still diagnostic challenge in differentiating FP and PDAC [2, 3]. In this work, we propose a novel method to address the need above with a new parameter (inhomogeneity index) based on the ADC map analysis with different region of interest (ROI) size.
Sixty-four patients with pathology-proven PDAC, six patients with pathology-proven FP and eighteen healthy volunteers were recruited and underwent DWI (b-values = 0, 600 s/mm2) on 3.0T whole body system (GE HDxt). ADC maps were calculated on a voxel-by-voxel basis from the obtained DWI images using mono-exponential model (ADC = [ln (SIb0 / SIb600)] / 600 ), and were reconstructed with FOV of 380*380 mm2 and matrix of 256*256. A homemade software was used to measure mean value and standard deviation of ADC (MeanADC and SDADC) within each of 12 concentric round ROIs (areas: 20, 42, 59, 82, 99, 121, 139, 161, 176, 196, 227, and 242 mm2 with pixel numbers: 9, 19, 27, 37, 45, 55, 63, 73, 80, 89, 103 and 110, respectively) drawn on the solid part of the mass of lesions and the head of normal pancreas on the single slice of the ADC map, which contained the largest available targeted tumor area or normal pancreas, as illustrated in Fig.1. The inhomogeneity index, defined as the ratio of SDADC over MeanADC in each ROI, was also calculated for the lesions and normal pancreas for each of the ROI with different sizes. Water phantom was used to calculate the ideal inhomogeneity index as a reference measurement.
The averaged inhomogeneity index curves of PDAC (64 cases), FP (8 cases), normal pancreas (18 cases) and water phantom (8 cases) are shown in Fig.2. Significant differences were observed for the inhomogeneity index measured by 12 different-size ROIs among FP, PDAC and normal pancreas. Near linear increment of the inhomogeneity index was observed on healthy volunteers with expanding ROI, while non-linear increasing character was observed for patients with FP and PDAC. A turning point of 55 mm2 was also seen: with region size below 55 mm2, inhomogeneity index of PDAC increases slower than FP; while with size above 55 mm2, inhomogeneity index of PDAC increases faster. Inhomogeneity index of PDAC is uniformly bigger than that of FP in areas detected.
It is well known that lesions of pancreas usually feature with inhomogeneity. ADC measurements have been utilized to investigate pancreatic diseases; however, few studies have detected the inhomogeneity properties derived from different ROI sizes in PDAC or FP. This is the first study revealing the variation tendency of inhomogeneity and the inhomogeneity index curve could be used, but not limited, in differentiating PDAC, FP and normal pancreas. This approach might be valuable for differentiation in other disease and could potentially be applied in computer aided diagnose. The pathological causes associated with ADC heterogeneity may be further studied to better understand its utility.