Different Machine Learning and Deep Learning Methods for the Classification of Colorectal Cancer Lymph Node
Jin Li1, Peng Wang1, Yang Zhou1,2
1College of Intelligent System Science and Engineering, Harbin Engineering University, Harbin, China.
Frontiers in Bioengineering and Biotechnology
|February 1, 2021
Summary
Deep transfer learning effectively classifies colorectal cancer (CRC) lymph node metastasis (LNM), outperforming other methods. This aids in better treatment planning for CRC patients.
Area of Science:
- Oncology
- Medical Imaging
- Artificial Intelligence
Background:
- Accurate classification of colorectal cancer (CRC) lymph node metastasis (LNM) is crucial for patient prognosis and treatment strategy.
- Current automated methods for CRC LNM classification lack established effectiveness, necessitating comparative analysis.
Purpose of the Study:
- To evaluate and compare the performance of machine learning, deep learning, and deep transfer learning for automated CRC LNM classification.
- To identify the most effective AI-driven approach for improving diagnostic accuracy in CRC LNM detection.
Main Methods:
- Utilized 3,364 CRC patient samples (1,646 positive, 1,718 negative) for analysis.
- Extracted global and local features for machine learning algorithms; raw image data for deep learning models.
- Employed deep transfer learning, deep learning, and eight machine learning algorithms for classification.
Main Results:
- Deep transfer learning achieved the highest accuracy (0.7583) and Area Under the Curve (0.7941).
- Classification heat-maps were utilized to enhance the interpretability of deep learning and deep transfer learning models.
Conclusions:
- Deep transfer learning demonstrates superior performance in automated CRC LNM classification compared to traditional machine learning and standard deep learning.
- The findings support the clinical adoption of deep transfer learning for enhanced CRC LNM detection and personalized treatment planning.
