HYBRID RESEARCH ON LIVER CANCER CLASSIFICATION AND SEGMENTATION USING ADVANCED ALGORITHMS

Liver cancer remains a major global health challenge, necessitating precise classification and segmentationtechniques for early diagnosis and effective treatment. This research presents a hybrid approach that integratesmachine learning and deep learning algorithms to enhance liver cancer detection, classification, and segmentation.The study leverages clinical data and medical imaging, combining XGBoost, Random Forest, SVM, and LogisticRegression for classification, alongside CNN, ResNet, and U-Net for segmentation. To optimize modelperformance, advanced hyperparameter tuning techniques such as GridSearchCV, RandomizedSearchCV, andBayesian Optimization are employed. The proposed hybrid model aims to improve accuracy, precision, and recallin detecting liver cancer by effectively handling heterogeneous data sources. Traditional methods often sufferfrom overfitting, poor generalization, and high computational complexity. Our approach mitigates theselimitations by leveraging ensemble learning and deep feature extraction, ensuring a robust and scalableframework. Extensive experiments are conducted on publicly available liver cancer datasets, evaluatingperformance based on accuracy, sensitivity, specificity, and Dice similarity coefficient (DSC). The resultsdemonstrate that the hybrid model outperforms conventional classification and segmentation techniques,providing more reliable and interpretable predictions. This research contributes to the field of medical imagingand predictive analytics, offering a novel methodology for early-stage liver cancer detection. The findings havesignificant implications for computer-aided diagnosis (CAD) systems, improving clinical decision-making andpatient outcomes. Future work includes extending the approach to multi-modal imaging and real-time diagnostics,ensuring wider applicability in personalized medicine and precision oncology.