A Decision Support System Based on Neural Network and Genetic Algorithm: Case Study of Breast Cancer

Artificial intelligence has been changingthe way healthcare has been provided in many high-risk environments or areas with poor healthcare facilities. The emergence of epidemics and new diseases, as well as the crucial role of early diagnosis in prevention and the adoption of more effective treatments have highlighted the need for accurate design and self-organization of Clinical Decision Support Systems (CDSSs).

In this study, a CDSS based on a neural networks (NN) and genetic algorithm is proposed. Since, on the one hand,the performance of the neural network (NN) is highly dependent on its parameters, and on the other hand, there is a constant need for optimization experts to fine-tune the parameters in the use of new data, a new chromosomal structure is proposed to automatically extract the optimal NN architecture, the number of layers and neurons. The goal is to increase the reusability of the model and ease of use by health experts.

To evaluate the performance of the model, two standard breast cancer (BC) datasets, WBC and WDBC, were used. The model uses 70% of the data set for training and the remaining equally used for evaluation and testing. The test accuracy of the proposed model on WBC and WDBC datasets was 98.51 and 97.55%, respectively. The optimal NN architecture on WBC consisted a three-hidden layers with 18, 15 and 19 neurons in each layers, and on WDBC consisted one hidden layer with a single neuron.

The proposed chromosomal structure is able to derive optimal NN architecture. In according to the high classification accuracy of the model in the diagnosis of BC and providing the different architectures in accordance with the hardware implementation considerations, the proposed model can be used effectively in the diagnosis of other diseases.