Thoracic lymph node station recognition on CT images based on automatic anatomy recognition with an optimal parent...

Area of Science:

• Medical Imaging
• Radiology
• Computer-Aided Diagnosis

Background:

• Accurate lymph node detection and segmentation in medical images, especially low-dose CT scans, remain challenging due to low contrast and anatomical variations.
• Existing methods struggle with the precise localization of lymph node stations, particularly those defined by the International Association for the Study of Lung Cancer (IASLC) lymph node map.

Purpose of the Study:

• To develop and validate an enhanced automatic anatomy recognition (AAR) framework for accurate localization of thoracic lymph node stations on computed tomography (CT) images.
• To integrate strategies for composite lymph node station definition and optimal hierarchical anchoring within the AAR framework.

Main Methods:

• Utilized a previously developed automatic anatomy recognition (AAR) framework, employing a one-shot method for localizing lymph node stations as anatomic objects.
• Implemented two integration strategies: grouping nearby lymph node stations into composite stations and establishing optimal organ-based anchors for each station.
• Trained and tested the framework using 28 contrast-enhanced thoracic CT datasets for model building and 12 independent datasets for validation.

Main Results:

• The AAR framework successfully localized thoracic lymph node stations with high precision, achieving accuracy within 2-3 voxels compared to ground truth.
• Demonstrated the effectiveness of the proposed strategies for composite station definition and hierarchical anchoring in improving localization accuracy.

Conclusions:

• The enhanced AAR framework provides a robust and accurate solution for thoracic lymph node station localization in CT imaging.
• This approach holds significant potential for improving cancer staging and treatment planning, particularly in challenging low-dose CT acquisitions.