A Transfer Learning-Based Framework for Classifying Lymph Node Metastasis in Prostate Cancer Patients.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Prostate cancer is a common diagnosis, with spread to lymph nodes indicating aggressive disease.
• Differentiating malignant from non-malignant lymph nodes on imaging is challenging for radiologists.
• Artificial intelligence (AI) offers potential solutions for medical imaging diagnostic tasks.

Purpose of the Study:

• To develop and evaluate a scalable hybrid AI framework for identifying malignant lymph nodes in prostate cancer patients.
• To compare the performance of the hybrid AI approach against traditional texture algorithms (GLCM, Gabor).

Main Methods:

• A hybrid framework utilizing a pre-trained deep learning model (ResNet-18) for feature extraction.
• Features extracted by ResNet-18 were fed into a machine learning classifier for lymph node identification.
• Comparison with classification models using Gray-Level Co-occurrence Matrix (GLCM) and Gabor texture features.

Main Results:

• The proposed hybrid framework achieved 76.19% accuracy, 79.76% sensitivity, and 69.05% specificity.
• Traditional texture algorithms showed lower performance: GLCM (61.90% accuracy) and Gabor (65.08% accuracy).
• The hybrid approach demonstrated superior performance in classifying prostate lymph nodes.

Conclusions:

• A hybrid AI approach using deep learning for feature extraction followed by machine learning classification is a viable solution.
• This hybrid method is particularly beneficial for medical imaging applications dealing with small datasets.
• The study highlights the potential of AI in improving the diagnosis of aggressive prostate cancer.