A Hybrid Deep Learning and Machine Learning Approach with Mobile-EfficientNet and Grey Wolf Optimizer for Lung and...

Area of Science:

• Computational pathology
• Artificial intelligence in oncology
• Medical image analysis

Background:

• Lung and colon cancers are leading causes of cancer mortality globally.
• Accurate histopathological classification is crucial for effective cancer diagnosis and treatment.
• Existing diagnostic methods require enhancement for improved generalizability and real-time application.

Purpose of the Study:

• To develop a hybrid deep learning and machine learning framework for classifying five types of histopathological images: Colon Adenocarcinoma, Colon Benign Tissue, Lung Adenocarcinoma, Lung Benign Tissue, and Lung Squamous Cell Carcinoma.
• To improve the generalizability of cancer classification models by addressing limitations in existing datasets.

Main Methods:

• Utilized Contrast Limited Adaptive Histogram Equalization (CLAHE) for image enhancement.
• Introduced new images from the National Cancer Institute GDC Data Portal to diversify the dataset.
• Developed a hybrid feature extraction model (MobileNetV2 and EfficientNetB3) optimized with Grey Wolf Optimizer (GWO).
• Applied machine learning models (XGBoost, LightGBM, CatBoost) with cross-validation and hyperparameter tuning using Optuna.

Main Results:

• The proposed MEGWO-LCCHC technique demonstrated high classification accuracy.
• The lightweight deep neural network (DNN) model achieved 94.8% accuracy.
• LightGBM, XGBoost, and CatBoost models achieved accuracies of 93.9%, 93.5%, and 93.3%, respectively.

Conclusions:

• The developed framework significantly enhances classification performance for lung and colon cancers.
• The approach offers improved generalizability, making it suitable for real-world clinical applications.
• The MEGWO-LCCHC framework shows potential as a robust tool for advancing AI in cancer diagnostics.