Enhancements in Radiological Detection of Metastatic Lymph Nodes Utilizing AI-Assisted Ultrasound Imaging Data and...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Oncology
• Diagnostic Pathology

Background:

• Accurate detection of neoplastic lesions in lymph nodes (LNs) is critical for cancer staging and prognosis.
• Current diagnostic methods can be time-consuming and may benefit from automation.
• The Lymph Node Reporting and Data System (LN-RADS) scale provides a standardized framework for LN assessment.

Purpose of the Study:

• To develop and evaluate a novel machine learning (ML) approach for the automatic detection of neoplastic lesions in lymph nodes.
• To investigate the impact of integrating the LN-RADS scale into ML models for improved diagnostic accuracy.
• To compare the performance of different ML architectures, including Multinominal Logistic Regression (MLR) and fully connected neuron layers.

Main Methods:

• Development of ML algorithms incorporating histopathological data and LN-RADS scale labels.
• Training and evaluation of models using three distinct combinations of data and labels.
• Comparative analysis of MLR-integrated models and fully connected neuron layer models.

Main Results:

• The integration of the LN-RADS scale significantly improved prediction accuracy in neoplastic lesion detection.
• MLR-integrated models demonstrated higher overall accuracy compared to other approaches.
• Fully connected neuron layer models achieved superior Area Under the Curve (AUC) performance.

Conclusions:

• AI techniques, particularly ML models combined with the LN-RADS scale, show significant potential for improving early cancer detection and prognosis.
• Further research into diverse datasets and network architectures can lead to enhanced diagnostic capabilities.
• The study highlights the utility of ML for automating and refining lymph node lesion analysis.