Improvement in tangential breast planning efficiency using a knowledge-based expert system

Abstract

A knowledge-based expert system was developed for the purpose of improving radiotherapy planning efficiency for a standardized, tangential breast technique. Treatment parameters pertaining to 150 previously planned patients were used for correlating the midplane breast contour of a new patient with an appropriate set of tangential beam weights and wedge angles; other treatment parameters including, planning target volume and isocenter, were specified by a radiation oncologist. Treatment plans generated by the expert system approach and a traditional, dosimetric approach were compared and rated prospectively in 45 patients. In addition, planning time was measured for both approaches. A performance rating of 97% was achieved for the expert system, in which an artificial neural network was used to correlate breast contours to treatment parameters, and approximately 30 minutes per patient was saved in treatment planning time. This high performance rating validated various assumptions concerning the expert system: namely, that the resultant dose distribution was not influenced by tangential field width (within the range of 7 to 12 cm), nominal beam energy (6 MV), or wedge type (physical vs. enhanced dynamic). Hence, the knowledge base may be directly transferable to other cancer centers using the same breast technique, and suggests that a global resource of radiotherapy treatment plans as well as planning strategies, categorized by treatment site, stage, and technique, may be viable.