Identification of Intra-abdominal Organs Using Deep Learning Techniques

Presently deep learning techniques are playing important role in making healthcare systems more intelligent, efficient, and effective. Proposed methodology is an advisory system in medical imaging which helps in clinical diagnosis. In medical imaging, ultrasound imaging is most frequently used as it is safe, painless, not exposed to ionizing radiation, and it allows real-time imaging. Ultrasound imaging takes more time in diagnosis and well-trained radiologist for interpreting and understanding. Hence, proposed system acts as an advisory system in identifying intra-abdominal organs and abnormalities if any. In this proposed system, the data was collected from intra-abdominal ultrasound images that do not contain any exploring information about the patient. Using filters, noise in ultrasound images is removed. Organ is segmented from ultrasound image and is identified by using deep neural network, and using shape and texture features, abnormalities are identified if any. At the end, various challenges that exist with deep neural network and ultrasound images are discussed.