Detection of Liver Tumour Using Deep Learning Based Segmentation with Coot Extreme Learning Model.

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Oncology

Background:

• Early liver cancer detection is critical for effective treatment.
• Liver tumors present segmentation challenges due to similar tissue characteristics and irregular shapes.
• Existing segmentation methods have limitations in accuracy and efficiency for liver tumor analysis.

Purpose of the Study:

• To develop an efficient deep learning approach for liver tumor segmentation and classification.
• To improve the accuracy of liver cancer detection using multimodal intelligence.
• To address the limitations of conventional and CNN-based interactive segmentation methods.

Main Methods:

• A novel deep learning segmentation approach combining geodesic distance encoding with user interaction.
• Utilizing a Coot Optimization Algorithm (COA) to optimize parameters for an Extreme Learning Model (ELM) classifier.
• Employing a deep learning-based Segmentation with Coot Extreme Learning Model (S-CELM) on the 3D-IRCADb1 dataset.

Main Results:

• The proposed S-CELM approach demonstrated high efficiency in segmenting liver tumors.
• Significant improvements in segmentation quality metrics (DICE and accuracy) were observed compared to existing methods.
• Accurate detection of tumors from publicly available liver image datasets was achieved.

Conclusions:

• The developed deep learning model offers a promising solution for accurate liver tumor segmentation and classification.
• This approach enhances medical analytics and decision-making in healthcare, particularly for liver cancer.
• The S-CELM method provides a more effective tool for early liver cancer identification and diagnosis.