A CAD system for detection of lung damage

Student thesis: Master thesis (including HD thesis)

  • Jens Tranholm Olesen
CAD systems are increasingly relied on for the identification and quantification of lung damage following radiation therapy. However the ability to classify several types of lung damage using the same features and classifier is often difficult. As a part of this work, the use of different subvolume sizes was examined, showing that larger volumes provide better classification. The use of sub volumes of size 21x21x21 to distinct between treated non-pneumonitis tissue and treated pneumonitis tissue, was found to produce an accuracy of 88.7%. However, future work should focus on including the counter lateral lungs in the classification.
LanguageEnglish
Publication date2012
ID: 63546695