Applications of Artificial Neural Network to Building Statistical Models for Qualifying and Indexing Radiation Treatment Plans

Abstract—The main goal in this paper is to quantify the quality of different techniques for radiation treatment plans, a back-propagation artificial neural network (ANN) combined with biomedicine theory was used to model thirteen dosimetric parameters and to calculate two dosimetric indices. The correlations between dosimetric indices and quality of life were extracted as the features and used in the ANN model to make decisions in the clinic. The simulation results show that a trained multilayer back-propagation neural network model can help a doctor accept or reject a plan efficiently. In addition, the models are flexible and whenever a new treatment technique enters the market, the feature variables simply need to be imported and the model re-trained for it to be ready for use.

Keywords—neural network, dosimetric index, radiation treatment, tumor

I. INTRODUCTION

OWADAYS, the combination of biomedicine and information engineering is a major branch of research in the world. Radiation treatment plays an important role in curing cancer. Almost all kinds of cancer can be treated by medical radiation. High-energy X-rays or gamma rays are delivered at the tumor; however radiation also harms the normal tissue around the target. Each organ has its own constraints of radiation doses. If the radiation dose to an organ exceeds its limit, this will cause permanent injury or even death. In the opposite case, radiation treatment is ineffective if the delivered dose is lower to prevent damage to normal organs. Therefore, the technique of delivering the desired dose to the target and reducing it to a reasonable level on normal organs has been improved in the history of radiation treatment. In our research, we wish to study biomedicine signals and analyse the patient database by medical engineering. The statistical data collected from a department of radiation oncology are analyzed and compared; this helps engineers edit the program. In addition, we also wish to adopt modern algorithms to simulate the relationship between normal tissues and cancer.

II. MATERIAL AND METHOD

In today’s treatment planning system, there are two commonly methods used to describe the distribution of radiation dose: the isodose curve and dose-volume histogram (DVH).

Authors are with Kaohsiung University of Applied Sciences, Taiwan. e-mail:[email protected]

However, if we wish to compare different treatment plans at the same time, the only way that this can be done is to evaluate plans based on a doctor’s experience. The disadvantages are the time taken and sometimes an improper decision being made. Fortunately, a neural network offers a faster and more precise way to compare treatment plans.

This research adopts the concept of dosimetric index and dosimetric parameters which include four evaluated parameters for treatment target, nine evaluated indices and two dosimetric parameters for normal tissue. In the next step, we choose three kinds of treatment plan to compare their quality with the specified dosimetric indices and parameters mentioned above. The first is a conventional seven-field intensity modulated radiotherapy (IMRT) plan. The next is an 18-field IMRT plan which offers a well-shaped dose distribution on the edge of targets but takes two to three times longer than the first. The latest technique which is so-called volumetric modulated arc therapy (VMAT) is the final plan. The neural network is well suited to constructing analytical modules because of its self-adaptive and powerful learning ability. Therefore, the calculated result of a neural network can be a reference as the doctor decides to reject or accept a treatment plan.

A. Dosimetric index and parameter

A case of nasopharyngeal cancer will be an example for describing the relationship between radiated volume of normal tissue and radiation dose around a tumor with some dosimetric indices and parameters. These indices and parameters are constructed by the neural network and its output is available when choosing a suitable conformity index and homogeneous index. These dosimetric parameters are defined as follows.

Dosimetric parameters: These parameters are obtained by

the range of dose distributions on a tumour. The term V93 means that at least 97 percent of the volume of the tumour accepts 93 percent of the prescribed dose. V100 means that more than 95 percent of the volume of the tumour is covered by the total prescribed dose. The volume of the tumour receiving 110 percent of the prescribed dose is less than 20% which is defined as V110. Similarly, less than five percent of volume is radiated by 115 percent of the prescribed dose which is called V115 [1-2].

Pei-Ju Chao, Tsair-Fwu Lee, Wei-Luen Huang, Long-Chang Chen, Te-Jen Su, and Wen-Ping Chen

Applications of Artificial Neural Network to

Building Statistical Models for Qualifying and

Indexing Radiation Treatment Plans

N

World Academy of Science, Engineering and Technology Vol:4 2010-09-22

734

also thankful for supplies from the National Science Council (NSC 98-221-E-151-038).

REFERENCES

[1] T. F. Lee, P. J. Chao, C. Y. Wang, J. H. Lan, Y. J. Huang, H. C. Hsu, C. C. Sung, T. J. Su, S. L. Lian, and F. M. Fang, "Dosimetric comparison of helical tomotherapy and dynamic conformal arc therapy in stereotactic radiosurgery for vestibular schwannomas," Medical Dosimetry, vol.

[2] T. F. Lee, F. M. Fang, P. J. Chao, T. J. Su, L. K. Wang, and S. W. Leung, "Dosimetric comparisons of helical tomotherapy and step-and-shoot intensity-modulated radiotherapy in nasopharyngeal carcinoma," Radiother Oncol, vol. 89, pp. 89-96, 2008.

[3] A. U. Khan, T. K. Bandopadhyaya, and S. Sharma, "Comparisons of Stock Rates Prediction Accuracy Using Different Technical Indicators with Backpropagation Neural Network and Genetic Algorithm Based Backpropagation Neural Network," Emerging Trends in Engineering and Technology, 2008. ICETET '08. First International Conference on, pp. 575-580, 2008.

[4] H. F. Soliman, A. M. Sharaf, M. M. Mansour, S. A. Kandil, and M. H. El-Shafii, "Adaptive ANN Rule-Based Controller for a Chopper

fed PMDC Motor Electric Vehicles Drive," Intelligent Vehicles '94 Symposium, Proceedings of the, pp. 429-434, 1994.

[5] A. Islam, M. R. Hasan, R. Rahaman, S. M. R. Kabir, and S. Ahmmed, "Designing ANN using sensitivity & hypothesis correlation testing " iccit 2007. 10th international conference on, Computer and information technology, pp. 1-6, 2007.

[6] M. Simsek and N. S. Sengor, "An Efficient Inverse Ann Modeling Approach Using Prior Knowledge Input with Difference Method," Circuit Theory and Design, 2009. ECCTD 2009. European Conference on, pp. 323-326, 2009.

[7] L. Xinran, W. Lide, and L. Peiqiang, "The Study on Composite Load Model Structure of Artificial Neural Network," Electric Utility Deregulation and Restructuring and Power Technologies, 2008. DRPT 2008. Third International Conference on, pp. 1564-1570, 2008.

[8] L. Jian-Chang, N. Dong-Xiao, and J. Zheng-Yuan, "A study of short-term load forecasting based on ARIMA-ANN," Proceedings of 2004 International Conference on, Machine Learning and Cybernetics, pp. 3183-3187, 2004.

[9] S. Kuan-Hao, W. Liang-Chih, L. Jih-Shian, L. Ren-Shyan, and C.

World Academy of Science, Engineering and Technology Vol:4 2010-09-22

738

Jyh-Cheng, "A Novel Method to Improve Image Quality for 2-D Small Animal PET Reconstruction by Correcting a Monte Carlo-Simulated System Matrix Using an Artificial Neural Network," IEEE Transactions on, Nuclear Science, vol. 56, pp. 704-714, 2009. [10] I. Masood and A. Hassan, "Synergistic-ANN Recognizers for

Monitoring and Diagnosis of Multivariate Process Shift Patterns," Soft Computing and Pattern Recognition, 2009. SOCPAR '09. International Conference of, pp. 266-271, 2009.

[11] M.-Y. Cho, T.-F. Lee, S.-W. Kau, C.-S. Shieh, and C.-J. Chou, "Fault diagnosis of power transformers using SVM/ANN with clonal selection algorithm for features and kernel parameters selection," 1st International Conference on Innovative Computing, Information and Control 2006, ICICIC'06, August 30, 2006 - September 1, 2006, Beijing, United states, pp. 26-30, 2006.

[12] H.-Y. Wu, C.-Y. Hsu, T.-F. Lee, and F.-M. Fang, "Improved SVM and ANN in incipient fault diagnosis of power transformers using clonal selection algorithms," International Journal of Innovative Computing, Information and Control, vol. 5, pp. 1959-1974, 2009. [13] K. Taji, T. Miyake, and H. Tamura, "On Error Backpropagation

Algorithm Using Absolute Error Function," 1999 IEEE International Conference on, Systems, Man, and Cybernetics, pp. 401-406, 1999. [14] M. A. Sovierzoski, F. I. M. Argoud, and F. M. de Azevedo,

"Evaluation of ANN Classifiers During Supervised Training with ROC Analysis and Cross Validation," BioMedical Engineering and Informatics, 2008. BMEI 2008. International Conference on, pp. 274-278, 2008.

[15] Y. Yoru, T. H. Karakoc, and A. Hepbasli, "Application of Artificial Neural Network (ANN) Method to Exergy Analysis of Thermodynamic Systems," Machine Learning and Applications, 2009. ICMLA '09. International Conference on, pp. 715-718, 2009.

World Academy of Science, Engineering and Technology Vol:4 2010-09-22

739