Computer Aided Diagnosis of Oral Cancer: Using Time-Step CT Images

摘要

The ability for medical professionals, and specifically oncologists, to look at and analyze radiographic images is severely limited and the need always exceeds the ability. Because there is a great need for fast reliable analysis of radiographic images, this project was started to create an algorithm for automatically searching images for a cancerous region. This algorithm looks at CT volumetric data sets; more specifically at the case of time-step CT images. This means that the first CT images was taken, analyzed by a medical professional and declared to be non-cancerous. Afterward, this algorithm can take future CT images and analyze them for signs of cancer. The algorithm created during this project first imports the volumetric data sets, and then registers the data sets. Once the images are registered, the images can be subtracted to find any ROIs. This is done by filtering the image by an area measure, and then by looking at a texture measure that calculates the uniformity or smoothness texture measures. When the whole algorithm was implemented, the end result was very successful.