Implementation of a deep convolution neural network model for identifying and classifying Pleuropulmonary Blastoma on DNA sequences

Abstract

Pleuro Pulmonary Blastoma is a lung cancer affecting children's lung tissues. The development of the DICER1 gene in the patients leads to PPB. This disease causes trouble breathing along with lung infections. It can only be found by analyzing the DNA sequence of the PPB tissues. The DNA sequence of the tissues affected by cancer is extracted from the infections. The DNA sequences reveal the hereditary cause of the PPB, which can help in detecting the abnormalities of their family members. DNA sequences are complex, and it is impossible to rely on manual clinical methods to give faster and more accurate predictions. Several research works have suggested machine learning models which process the DNA sequences to detect PPB. However, machine learning algorithms cannot understand the complex DNA patterns of cancer tissues. This paper designs and implements a deep learning algorithm, a novel Deep Convolution Neural Network, to overcome such difficulties. The whole DNA sequence can be processed by more layers involved in the DCNN. The proposed DCNN learns the DNA sequence data part by part and compares it with the ground truth values, identifying odd data and abnormal sequence patterns to predict the presence of PPB. The DCNN is implemented in Python software with the DICER1 gene dataset for predicting the PPB. From the result of the experiment, the efficacy of the proposed DCNN model is verified and compared with other traditional models to evaluate the performance. The proposed CNN obtained 98.67% of accuracy for CT image classification and 96% of accuracy for DICER1-DNA data analysis.

Keywords

• Pleuropulmonary Blastoma
• Lung cancer
• Cancer diagnosis
• Medical image processing