東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Interstitial brachytherapy cancer tr...

Miller, Steven Robert Guy.

FindBook

Google Book

Amazon

博客來

Interstitial brachytherapy cancer treatment optimization using simulated annealing and artificial neural networks.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Interstitial brachytherapy cancer treatment optimization using simulated annealing and artificial neural networks./
作者:	Miller, Steven Robert Guy.
面頁冊數:	501 p.
附註:	Source: Masters Abstracts International, Volume: 42-01, page: 0277.
Contained By:	Masters Abstracts International42-01.
標題:	Engineering, Biomedical. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=MQ79986
ISBN:	0612799867

Interstitial brachytherapy cancer treatment optimization using simulated annealing and artificial neural networks.
Miller, Steven Robert Guy.

Interstitial brachytherapy cancer treatment optimization using simulated annealing and artificial neural networks. - 501 p.

Source: Masters Abstracts International, Volume: 42-01, page: 0277.

Thesis (M.Sc.)--The University of Manitoba (Canada), 2003.

Optimization of interstitial brachytherapy implants has recently turned to non-deterministic optimization techniques, such as simulated annealing (SA) and genetic algorithms (GA). However, the current SA and GA approaches have three major limitations: (i) they are computationally expensive, with the fastest being reported at 3 minutes of dedicated CPU time for a single solution, (ii) they are limited to evaluating seed positions at predefined needle positions, and (iii) they can not be used to update plans during needle insertion. In order to address these shortcomings, a system has been designed and implemented which uses SA and an artificial neural network (ANN). The role of the SA is to find optimal source placements within a tumour from which the ANN can be trained. If the training of the ANN is carried out properly, it is able to generalize the training data, and is capable of computing optimized brachytherapy cancer treatments in milliseconds.

ISBN: 0612799867Subjects--Topical Terms:

1017684
Engineering, Biomedical.

Interstitial brachytherapy cancer treatment optimization using simulated annealing and artificial neural networks.
LDR:02834nmm 2200301 4500 001 1862514
005 20041202093454.5
008 130614s2003 eng d
020 $a 0612799867
035 $a (UnM)AAIMQ79986
035 $a AAIMQ79986
040 $a UnM $c UnM
100 1 $a Miller, Steven Robert Guy. $3 1950063
245 1 0 $a Interstitial brachytherapy cancer treatment optimization using simulated annealing and artificial neural networks.
300 $a 501 p.
500 $a Source: Masters Abstracts International, Volume: 42-01, page: 0277.
500 $a Adviser: W. Kinsner.
502 $a Thesis (M.Sc.)--The University of Manitoba (Canada), 2003.
520 $a Optimization of interstitial brachytherapy implants has recently turned to non-deterministic optimization techniques, such as simulated annealing (SA) and genetic algorithms (GA). However, the current SA and GA approaches have three major limitations: (i) they are computationally expensive, with the fastest being reported at 3 minutes of dedicated CPU time for a single solution, (ii) they are limited to evaluating seed positions at predefined needle positions, and (iii) they can not be used to update plans during needle insertion. In order to address these shortcomings, a system has been designed and implemented which uses SA and an artificial neural network (ANN). The role of the SA is to find optimal source placements within a tumour from which the ANN can be trained. If the training of the ANN is carried out properly, it is able to generalize the training data, and is capable of computing optimized brachytherapy cancer treatments in milliseconds.
520 $a The system developed in this thesis is the first step towards an ANN-based optimization technique for interstitial brachytherapy. The SA is designed to optimize source placement within 2D tumour shapes and produces results that meet the requirements identified by a suitable cost function. The ANN component is designed to generate relative-dose distributions for 2D square tumour shapes using constant source strengths. Through experimentation, it has been determined that the most appropriate structure for the single hidden layer ANN has 12 interior nodes for tumours up to 3 cm in cross sectional size. Using this network layout, the ANN is able to achieve a root mean square (RMS) error of 2.03% of the relative dose on the final pass through the training data, an RMS error of 13.37% on a test set, an average positional error of 1.07 mm and a maximum of 3 mm in positional error, compared to the results created by the SA.
590 $a School code: 0303.
650 4 $a Engineering, Biomedical. $3 1017684
650 4 $a Health Sciences, Oncology. $3 1018566
650 4 $a Artificial Intelligence. $3 769149
690 $a 0541
690 $a 0992
690 $a 0800
710 2 0 $a The University of Manitoba (Canada). $3 1249364
773 0 $t Masters Abstracts International $g 42-01.
790 1 0 $a Kinsner, W., $e advisor
790 $a 0303
791 $a M.Sc.
792 $a 2003
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=MQ79986