Development and Validation of Machine Learning Model Platelet Index-based Predictor for Colorectal Cancer Stage

Abstract

INTRODUCTION: Colorectal cancer (CRC) staging is essential for effective treatment planning and prognosis. While platelet indices have shown promise in indicating CRC aggressiveness, a platelet index-based predictor for CRC staging has not been established in Indonesia. This study aimed to explore the relationship between platelet indices and CRC stage and to develop a predictive model and application. METHODS: This cross-sectional study analyzed 369 CRC patients from Dr. Wahidin Sudirohusodo Hospital. Key parameters included age, gender, tumor location, and platelet indices: platelet count (PC), mean platelet volume (MPV), platelet distribution width (PDW), plateletcrit, and the MPV/PC ratio. Data were processed using SPSS 25, MATLAB, and Streamlit. RESULTS AND DISCUSSION: The analysis revealed significant correlations between elevated platelet indices and advanced CRC stages. Various machine learning models were developed, with Support Vector Machine (SVM) achieving the highest accuracy at 82.9%, followed closely by K-Nearest Neighbors (82.7%), Neural Network (81.5%), Naive Bayes (80.5%), and logistic regression (51.5%). The most effective model was implemented as a portable application through Streamlit, yielding 79.2% internal validation and 89.2% external validation. CONCLUSION: This study highlights a significant association between increased platelet indices and advanced CRC stages. The innovative platelet index-based predictor for CRC staging offers promising potential for enhancing individualized clinical decision-making. By providing a non-invasive method that complements existing staging techniques, this approach could significantly improve patient outcomes through earlier and more accurate CRC staging. The findings underscore the importance of integrating simple, accessible biomarkers into clinical practice to enhance diagnostic precision.

Access to Document

Other files and links

• Link to publication in Scopus
• Open Access Version Available

Fingerprint