Diagnosis of leukemia using microarray analysis based on Hidden Markov Model and Random Convolutional Kernel Transform

Abstract

Introduction

Leukemia is one of the most prevalent cancers worldwide, and early detection is critical for effective treatment. Microarray data is a key tool in this process, given the vast number of genes involved, which makes the analysis complex and time-consuming. Identifying relevant genes is a crucial step in disease diagnosis.

Material and methods

This study aims to improve the diagnostic accuracy of various leukemia types by using microarray data in combination with advanced deep learning techniques. The proposed model begins with selecting essential features and sequences relevant to diagnosis. These data sequences are processed using a Generative Adversarial Network (GAN) with a U-Net architecture to generate synthetic data. Both the synthetic and original data are then labeled for analysis. Feature ranking is conducted using a Hidden Markov Model (HMM), followed by classification using the Random Convolutional Kernel Transformation (ROCKET) approach. This process ultimately predicts five leukemia categories within the sample.

Results

The results demonstrate that the proposed model achieves a high classification accuracy of 99.26 %, outperforming existing methods.

Conclusion

This research highlights the importance of leveraging DNA alterations associated with genetic mutations to improve leukemia diagnostics, emphasizing the potential for early detection and intervention. In simpler terms, identifying DNA modifications across the genome can help predict an individual's likelihood of developing leukemia. Detecting these changes can significantly aid in diagnosis.