DIFFERENTIATION BETWEEN PANCREATIC CYSTIC LESIONS USING IMAGE PROCESSING SOFTWARE (FIJI) BY ANALYZING ENDOSCOPIC ULTRASONOGRAPHIC (EUS) IMAGES

Aims EUS is the most accurate imaging modality for evaluation of different types of pancreatic cystic lesions; however, distinguishing between malignant and benign lesions remains challenging. Our aim was to analyze EUS images of pancreatic cystic lesions using an image processing software (FIJI).

Methods We specified echogenicity of the lesions by measuring the gray value of pixels inside the selected areas. Besides the entire lesion, its cystic and solid parts were also separately selected for assessment. Following the software analyzing process images were divided into groups (serous cystic neoplasm/SCN/, non-SCN and pseudocyst) according to the cytology results of the lesions. Intraductal papillary mucinous neoplasms (IPMNs) and mucinous cystic neoplasms (MCNs) were classified as non-SCN category.

Results EUS images of 33 patients (21 females, 12 males; mean age of 60.9 ± 10.1 and 66.3 ± 11.6 years, respectively) were assessed. Overall 73 images were processed by the software: 36 in non-SCN, 13 in SCN and 24 in the pseudocyst group. The mean gray value of the entire lesion in non-SCN group was significantly higher than in SCN group (31.7 vs 25.5; p = 0.022). The area ratio (area of cystic part/entire lesion) in non-SCN, SCN and pseudocyst group was 42%, 55% and 70%, respectively; significantly lower in non-SCN group than in SCN and pseudocyst group (p = 0.0058 and p < 0.0005, respectively). The lesion density (sum of the gray values/area of the lesion) was also significantly higher in non-SCN group compared to the SCN- and pseudocyst group (4802.48/mm² vs 3865.87/mm² vs 3192.27/mm²; p = 0.022 and p = 0.004, respectively). No correlation was found between the intracystic CEA levels and the analyzed cystic gray values.

Conclusions The computer-aided diagnosis decision is being used increasingly due to the rapid development of the information technology. The EUS image analysis process may have a potential to be a diagnostic tool for the evaluation and differentiation of pancreatic cystic lesions.