Stage-based colorectal cancer prediction on uncertain dataset using rough computing and LSTM models

Abstract

Abstract Artificial intelligence (AI) is an attractive field of Computer Science that helps to classify and to predict various real-time applications. Perhaps AI has a major role in predicting diseases at an early stage based on history. As cancer is one of the most harmful diseases where the mortality rate is high, it is now essential to utilize the benefits of AI to have an early diagnosis of cancer. Among various cancers, Colorectal cancer (CRC) is a common form of gastrointestinal cancer, and its treatment is lengthy and costly, with a high recurrence rate and high fatality rate. Initial disease analysis and prognosis are required to improve the patient’s treatment with a better survival analysis. However, the disease prediction process depends on the collected data, where the data may contain uncertainty. Uncertain data leads to wrong predictions. Thus, it is essential to utilize rough computing, a mathematical tool to deal with uncertainty. This paper has made an effort, to handle uncertainty using a rough set of fuzzy approximation space as pre-processing and utilized Unidirectional and Bidirectional LSTM for the classification and prediction process. Thus, to demonstrate improved predictive accuracy, the proposed model adapted the optimizers and evaluated using benchmarking techniques in predicting stage-based survival rate. The comparative analysis shows that the proposed model performs well against the state-of-the-art models and can help the medical practitioner to detect CRC at an early stage and reduce the mortality rate among human beings.

Keywords

• Colorectal cancer prediction
• Rough set on fuzzy approximation space (RSFAS)
• LSTM
• Survival analysis
• Weibull distribution