| Abstract [eng] |
Pancreatic cancer is a lethal disease that is hard to detect in the early stages. Poor outcomes of pancreatic cancer make the scientific community look for ways to improve cancer detection time and potentially save patients’ lives. Automating tumor detection is one of the ways to achieve this. Whileradiologicalimageclassificationcanmaketheprocesscheaperandfaster,imagesegmentation is important for detection of tumor edges, which aids diagnosis and treatment. One of the ways to detect pancreatic tumors in radiological images is deep learning. UNet is baseline standard for medical image segmentation and so far, performs very well on most tasks, though not so well on pancreatic cancer segmentation. Another way of research would be to try various computer vision algorithms to detect keypoints in pancreatic cancer computed tomography (CT) images, which can be used in conjunction with various machine learning algorithms. The ways to do it could be using computer vision methods like Scale Invariant Feature Transform (SIFT), Speeded Up Robust Feature (SURF), Features from Accelerated Segment Test (FAST), Oriented FAST and Rotated BRIEF (ORB), Binary Robust invariant scalable keypoints (BRISK), KAZE, Accelerated KAZE (AKAZE) or Adaptive and Generic Corner Detection Based on the Accelerated Segment Test (AGAST). Although there is some researchontheusageofthesemethodsfordetectingkeypointsinmedicalimages, itisnotextensive and none of it was applied to pancreatic cancer. This work attempts to try out computer vision algorithms which can find the keypoints corresponding to tumor in pancreatic cancer CT images and apply them to the images of Medical Segmentation Decathlon dataset. |
