Thyroid cancer is one of the fastest growing cancer diagnoses worldwide. Thyroid cancer commences from a typical growth of thyroid tissue at the edge of the thyroid gland. Thyroid image segmentation is one of the major steps in image analysis that subdivides an image into various constituent parts. The majority of image segmentation methods are based on thresholds obtained from image histogram. The difficulty in subdividing the image based on histogram is finding optimal threshold. Due to the complexity associated with histogram of thyroid medical images, the classical segmentation methods finds it difficult to obtain optimal threshold, thus either Swarm Algorithm (SA) or Evolutionary Algorithms (EA) are chosen to be an alternatives. The Backtracking Search Algorithm (BSA) is a newly introduced EA algorithm. The BSA has been proved to be very successful on various standard benchmark optimization problems and has one tuning parameter called amplitude control factor. This factor decides about generating trial population in BSA algorithm. Though there is only one control parameter, the trial population may be generated by various means centered on random strategy. This research paper develops various BSA variants based on trial population generation strategies and used for thyroid image segmentation. The comprehensive analysis of BSA variants is presented on multi-level image thresholding. It is found that BSA variants shows almost comparable solution and proves the robustness of BSA algorithm.