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

Area of Science:

• Computer Science
• Medical Informatics
• Data Science

Background:

• Colorectal cancer (CRC) poses significant health risks with high mortality and recurrence rates.
• Early diagnosis and prognosis are crucial for improving CRC patient outcomes and survival analysis.
• Data uncertainty in medical predictions can lead to inaccurate diagnoses and treatment plans.

Purpose of the Study:

• To develop an artificial intelligence (AI) model for early colorectal cancer (CRC) detection and survival rate prediction.
• To address data uncertainty in medical datasets using rough computing techniques.
• To enhance the accuracy of CRC prediction models for improved patient management.

Main Methods:

• Utilized rough set theory and fuzzy approximation spaces for data pre-processing to handle uncertainty.
• Employed Unidirectional and Bidirectional Long Short-Term Memory (LSTM) networks for classification and prediction.
• Adapted optimizers and evaluated the model using benchmarking techniques for stage-based survival rate prediction.

Main Results:

• The proposed AI model demonstrated improved predictive accuracy in identifying CRC stages and survival rates.
• Comparative analysis showed superior performance against existing state-of-the-art models.
• The model effectively handled data uncertainty through rough set fuzzy approximation pre-processing.

Conclusions:

• The developed AI model shows significant potential for early CRC detection, aiding medical practitioners.
• Addressing data uncertainty is critical for reliable disease prediction and improved patient outcomes.
• This approach can contribute to reducing the mortality rate associated with colorectal cancer.