Accurate, robust and reproducible segmentation of positron emission tomography (PET) images is very important
both in terms of radiotherapy planning as well as treatment assessment. However, due to high noise and poor resolution
of PET scanner, it still remains a daunting task. Threshold based segmentation methods are proved to be robust to
noise and resolution compared to other segmentation methods (e.g. clustering, gradient based etc.). Numerous publications
deal with the topic. This paper reviews different fixed and adaptive threshold methods proposed in literature in a
common mathematical framework. The paper also highlights the purpose of segmentation of PET images from two perspectives
– radiotherapy planning and treatment response. A few recommendations are suggested at the end.