Enhancing gastric cancer early detection: A multi-verse optimized feature selection model with crossover-information...

Area of Science:

• Oncology
• Bioinformatics
• Computational Medicine

Background:

• Early detection of gastric cancer (GC) is crucial for improving patient survival and treatment outcomes.
• Current diagnostic methods may have limitations in identifying nascent stages of the disease.
• Developing advanced screening tools is essential for timely intervention.

Purpose of the Study:

• To introduce and validate an innovative wrapper-based feature selection methodology, bCIFMVO-FKNN-FS, for early-stage GC detection.
• To enhance the accuracy and efficiency of identifying patients with early-stage GC.
• To identify key clinical parameters associated with early-stage GC.

Main Methods:

• Developed a novel bCIFMVO-FKNN-FS model integrating crossover-information feedback multi-verse optimizer (CIFMVO) with fuzzy k-nearest neighbors (FKNN).
• Validated CIFMVO performance on IEEE CEC benchmark functions against eleven state-of-the-art algorithms.
• Applied the bCIFMVO-FKNN-FS model to clinical data of 1632 patients diagnosed with early-stage GC or chronic gastritis.

Main Results:

• The CIFMVO algorithm demonstrated competitive optimization efficiency across various dimensionalities.
• The bCIFMVO-FKNN-FS model achieved a predictive accuracy of 83.395% and a sensitivity of 87.538% for early-stage GC detection.
• Identified age, gender, serum gastrin-17, serum pepsinogen I, and the PGI/PGII ratio as significant indicators for early-stage GC.

Conclusions:

• The proposed bCIFMVO-FKNN-FS model is effective for the early screening of gastric cancer.
• The identified clinical parameters provide valuable insights for early GC diagnosis.
• This research contributes to the development of advanced tools for timely GC detection and management.