ROENTGEN: a case-based radiation treatment planner
Jeffrey T. Berger · 1995
This dissertation describes an artificial intelligence approach to the problem of radiation treatment planning. In particular, it describes a case-based reasoning (CBR) system, Roentgen, that uses past treatment plans and past repair episodes to arrive at plans for new patients. While radiation treatment planning (RTP) is a domain that is inherently geometric, most previous CBR efforts deal with domains that are more symbolic and qualitative in nature. Design repair is a significant issue in RTP since the first attempt to produce a good plan almost always falls short of what is desired and the improvement of even an almost acceptable treatment plan so that it is acceptable can be a conditional and iterative process. The character of the interaction between the human and computer is important since the well-being of the patient demands that a human expert be embedded in the design loop. Hence, the main issues examined in this dissertation include indexing, geometric reasoning, design repair and human/computer interaction.