Knowledge-based enhancement of megavoltage images in radiation therapy using a hybrid neuro-fuzzy system

Megavoltage images (MVIs) are used in radiation therapy for verification of the patient's position during cancer treatment. Due to the physics of imag ing devices, the quality of MVI is very poor. In this work, we propose a hy brid neuro-fuzzy system consisting of fuzzy techniques and neural nets for knowledge-based enhancement of MVIs. The fuzzy enhancement includes differe nt contrast adaptation techniques and also soft filtering, respectively. A modified associative memory is trained using a priori knowledge for image r estoration. In order to consider the subjective demands of physicians, an o bserver-dependent overall system for contrast adaptation is also proposed. (C) 2001 Elsevier Science B.V. All rights reserved.