EE Seminar: AUTOMATIC DIAGNOSIS OF SLEEP DISORDERED BREATHING BASED ON LATERAL AND FRONTAL CEPHALOMETRIC IMAGES

~~Speaker: Orna Weiss, 
M.Sc. student under the supervision of Prof. Shai Avidan and Prof. Lior Wolf

Monday, July 27, 2015 at 15:30
Room 011, Kitot Bldg., Faculty of Engineering

AUTOMATIC DIAGNOSIS OF SLEEP DISORDERED BREATHING BASED ON LATERAL AND FRONTAL CEPHALOMETRIC IMAGES

Abstract

The aim of our project is to automatically analyze the lateral and frontal cephalometric images as a diagnostic tool for Sleep Disordered Breathing (SDB). In addition to the automatic algorithm that we develop, we also work on setting the gold standard in the field of SDB diagnosis. This includes establishing a standard terminology to analyze craniofacial structure anomalies from different imaging systems, such as CT and MRI, with an emphasis on cephalometric images. Our main focus is SDB diagnosis; however, many of the measurements and concepts we suggest can be used in other fields such as orthodontics.
Our automatic algorithm consists of a few stages. First, multiple landmarks in the cephalometric images are being registered. The landmark localization is based on a new registration method derived from Local Affine Frames (LAF). Multiple LAFs are sampled per image based on the random selection of triplets of key points, and are used to find the matching registration between the training and test images. The landmarks are then projected, using affine transformation, from the training images to the query image.
Next, anatomical measurements (e.g. distances, planes and angles) and structures (e.g. tongue, velum pharynx etc.) that were previously associated with SDB are localized.
Using the data we acquire, we can predict the severity of the disease. We can also identify the specific source of the problem and recommend the proper treatment.