Angiogenesis is a crucial event for tumor growth and it is regulated predominantly by several different growth factors. Vascular endothelial growth factor protein family (VEGF) and its receptors are probably the most important tissue factors responsible for angioblast differentiation and tube formation. VEGF protein family currently comprises several members: VEGF (or VEGF-A), VEGF-B, VEGF-C and VEGF-D, VEGF-F, placental growth factor (PlGF), and their receptors VEGFR-1, VEGFR-2 and VEGFR-3. VEGF is a key angiogenic growth factor and its level of expression is a critical marker for detection of the angiogenic diseases. The potent role of VEGF in tumor angiogenesis has been widely described in the past decade, being expressed in most types of nondigestive and digestive cancers. VEGF family members play an important role in the development of pancreatic cancer (especially VEGF-A, VEGF-C, VEGF-D, VEGFR-1 and VEGFR-2). VEGF-A is the most specific and prominent angiogenic factor among all family members and VEGFR-2 is the most important receptor in evaluating the angiogenesis in pancreatic cancer. Thus, VEGF overexpression may be considered as a diagnostic marker and as a poor prognostic factor of the disease.

Aim: It is well known that endoscopic ultrasound guided fine needle aspiration (EUS-FNA) has a high sensitivity (over 85%) and specificity (100%) for diagnosis of pancreatic cancer. The aim of the study was to establish a EUS based clinical diagnostic algorithm in patients with pancreatic masses and negative cytopathology after EUS-FNA, based on previously published results and cut-offs of real-time elastographic (RTE) EUS and contrast-enhanced harmonic (CEH) EUS. Material and methods: We included in the study a subgroup of 50 consecutive patients with focal pancreatic masses which underwent EUS examinations with negative EUS-FNA. RTE-EUS and CEH-EUS were performed sequentially in all patients. The sensitivity, specificity and accuracy of these methods were calculated separately. A clinical decision algorithm based on elastography followed by CEH was established. Results: For the diagnosis of possible malignancy, the sensitivity, specificity and accuracy of RTE-EUS were: 97.7%, 77.4%, and 84% respectively. CEH-EUS had similar results: 89.5%, 80.7%, and 84%, respectively. In 25 patients with soft/mixed appearance during elastography,sequential assessment using contrast-enhanced EUSwas performed. The specificity of CEH-EUS for detection of chronic pancreatitis in this sub-set of patients was excellent (100%). In other 25 patients with hard appearance in elastography (low strain) CEH-EUS had an excellent specificity (100%) and accuracy (93%) in the detection of pancreatic cancer. Conclusions: The proposed algorithm with sequential use of elastography followed by CEH could be a good clinical tool in the set of patients with negative EUS-FNA results for the differentiation between benign and malignant focal pancreatic masses.

Differential diagnosis of focal pancreatic masses is based on endoscopic ultrasound (EUS) guided fine needle aspiration biopsy (EUS-FNA/FNB). Several imaging techniques (i.e. gray-scale, color Doppler, contrast-enhancement and elastography) are used for differential diagnosis. However, diagnosis remains highly operator dependent. To address this problem, machine learning algorithms (MLA) can generate an automatic computer-aided diagnosis (CAD) by analyzing a large number of clinical images in real-time. We aimed to develop a MLA to characterize focal pancreatic masses during the EUS procedure. The study included 65 patients with focal pancreatic masses, with 20 EUS images selected from each patient (grayscale, color Doppler, arterial and venous phase contrast-enhancement and elastography). Images were classified based on cytopathology exam as: chronic pseudotumoral pancreatitis (CPP), neuroendocrine tumor (PNET) and ductal adenocarcinoma (PDAC). The MLA is based on a deep learning method which combines convolutional (CNN) and long short-term memory (LSTM) neural networks. 2688 images were used for training and 672 images for testing the deep learning models. The CNN was developed to identify the discriminative features of images, while a LSTM neural network was used to extract the dependencies between images. The model predicted the clinical diagnosis with an area under curve index of 0.98 and an overall accuracy of 98.26%. The negative (NPV) and positive (PPV) predictive values and the corresponding 95% confidential intervals (CI) are 96.7%, [94.5, 98.9] and 98.1%, [96.81, 99.4] for PDAC, 96.5%, [94.1, 98.8], and 99.7%, [99.3, 100] for CPP, and 98.9%, [97.5, 100] and 98.3%, [97.1, 99.4] for PNET. Following further validation on a independent test cohort, this method could become an efficient CAD tool to differentiate focal pancreatic masses in real-time.

No abstract