Prediction of survival in thyroid cancer using data mining technique.
Technol Cancer Res Treat
; 13(4): 353-9, 2014 Aug.
Article
en En
| MEDLINE
| ID: mdl-24206207
Cancer is the second leading cause of death after cardiovascular diseases in the world. Health professionals are seeking ways for suitable treatment and quality of care in these groups of patients. Survival prediction is important for both physicians and patients in order to choose the best way of management. Artificial Neural Network (ANN) is one of the most efficient data mining methods. This technique is able to evaluate the relationship between different variables spontaneously without any prevalent data. In our study ANN and Logistic Regression were used to predict survival in thyroid cancer and compare these results. SEER (Surveillance, Epidemiology and End Result) data were got from SEER site1. Effective features in thyroid cancer have been selected based on supervision by radiation oncologists and evidence. After data pruning 7706 samples were studied with 16 attributes. Multi Layer Prediction (MLP) was used as the chosen neural network and survival was predicted for 1-, 3- and 5-years. Accuracy, sensitivity and specificity were parameters to evaluate the model. The results of MLP and Logistic Regression models for one year are defined as for 1-year (92.9%, 92.8, 93%), (81.2%, 88.9%, 72.5%), for 3-year as (85.1%, 87.8%, 82.8%), (88.6%, 90.2%, 87.2%) and for 5-year as (86.8%, 96%, 74.3%), (90.7%, 95.9%, 83.7) respectively. According to our results ANN could efficiently represent a suitable method of survival prediction in thyroid cancer patients and the results were comparable with statistical models.
Texto completo:
1
Colección:
01-internacional
Base de datos:
MEDLINE
Asunto principal:
Neoplasias de la Tiroides
/
Redes Neurales de la Computación
/
Minería de Datos
Tipo de estudio:
Etiology_studies
/
Prognostic_studies
/
Risk_factors_studies
Límite:
Adult
/
Aged
/
Female
/
Humans
/
Male
/
Middle aged
Idioma:
En
Revista:
Technol Cancer Res Treat
Asunto de la revista:
NEOPLASIAS
/
TERAPEUTICA
Año:
2014
Tipo del documento:
Article
País de afiliación:
Irán
Pais de publicación:
Estados Unidos