Application of deep learning neural networks on Computed Tomography images of canine hepatic and splenic masses

Short description:

Hepatic and splenic masses are encountered frequently in dogs and reflect a range of pathological conditions including malignant neoplasia such as hepatocellular carcinoma and hemangiosarcoma, benign neoplasia such as hepatocellular adenoma and nonneoplastic conditions such as hepatic nodular hyperplasia and hematoma. Ability to noninvasively differentiate malignant from non malignant masses would aid clinical decision making. In recent works multiphase computed tomographic (CT) protocols have been used to examine the pattern of enhancement of hepatic lesions following administration of intravenous contrast medium. However, differentiation of malignant and nonmalignant hepatic lesions using contrast uptake characteristics resulted complicated because nonmalignant lesions (such as hepatic nodular hyperplasia) may also show signs of early marked contrast uptake followed by reduced uptake in later images, which mimics the appearance of malignant neoplasms. A specialised class of deep-learning architectures, the so-called convolutional neural networks, are considered the state-of-the-art algorithms for image analysis and classification; a substantial number of different applications are being developed in medical imaging for computer aided diagnosis.The aim of this PhD project is to develop a clinically applicable test based on deep neural networks to predict the histopathological diagnosis of canine hepatic/splenic masses through analysis of CT images.

Five publications related to the Research Topic for the interview:

  1. Dreyer KJ, Geis JR. When Machines Think: Radiology’s Next Frontier. Radiology. 2017;285(3):713–718
  2. Lakhani P, Sundaram B. Deep Learning at Chest Radiography: Automated Classification of Pulmonary Tuberculosis by Using Convolutional Neural Networks. Radiology. 2017;284(2):574–582
  3. Yasaka K, Akai H, Abe O, Kiryu S. Deep Learning with Convolutional Neural Network for Differentiation of Liver Masses at Dynamic Contrast-enhanced CT: A Preliminary Study. Radiology. 2017;286(3):17070
  4. Meng D, Zhang L, Cao G, Cao W, Zhang G, Hu B. Liver fibrosis classification based on transfer learning and FCNet for ultrasound images. IEEE Access. 2017;5(1):1–1
  5. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. 2015;521(7553):436–444