Affordable Access

Dose distribution studies of rectal cancer patients treated with brachytherapy

McGill University
Publication Date
  • Health Sciences - Radiology
  • Computer Science
  • Design
  • Medicine


We studied 42 rectal cancer brachytherapy patients with shielding included in the treatment, but without shielding included in the treatment planning and dose distribution calculations. To study the dose distribution in shielded cases, we first extracted relevant information from the PLATO BPS system and reconstructed the PLATO treatment plans with negligible errors. Based on this, we then implemented and validated the dose-superposition algorithm and applied this algorithm to all patients, finding that although the dose to healthy tissue is reduced, 10 of the 42 patients ended up with more than a 5% 'cold spot,' which is not acceptable in clinical practice. To solve this problem, a highly automated simulated annealing (SA) optimization algorithm, with a well-developed new cost function, was designed and implemented. The algorithm alone can decrease 'V90 to healthy tissue' by 17% while keeping the dose to target at the same level. By adding a shielding rod, the algorithm can further decrease the 'V90 to healthy tissue' by an additional 13%. The average calculation time for a patient using the dose-superposition algorithm was less than 20 seconds, and the time needed to make an SA based treatment plan for a shielded case was less than 10 minutes. The SA algorithm is independent of the physicist's experience. Two simplified algorithms for dose distribution calculation were investigated, the TG43 analytical shielding algorithm and the PLATO shielding algorithm. Their calculated results show conformal isodose lines to that from the dose-kernel-superposition algorithm, but present less accuracy and precision and much longer calculation times. Hence they are useful in limited conditions. Finally, we did some preliminary investigation

There are no comments yet on this publication. Be the first to share your thoughts.


Seen <100 times