Enhancing the Accuracy of Lymph-Node-Metastasis Prediction in Gynecologic Malignancies Using Multimodal Federated...

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Lymph node metastasis detection in gynecological malignancies is challenging for traditional imaging (CT, MRI, PET/CT).
• A diagnostic gap exists, necessitating innovative approaches for accurate metastasis identification.

Purpose of the Study:

• To bridge the diagnostic gap in gynecological cancer lymph node metastasis detection.
• To develop and evaluate a multimodal federated-learning model integrating non-image data and MRI analysis.

Main Methods:

• Developed a federated-learning framework using a composite neural network for diverse data integration.
• Employed a deep convolutional neural network with an enhanced U-NET architecture for MRI image processing.
• Compared the federated-learning model's performance with and without MRI data against traditional imaging.

Main Results:

• Traditional imaging showed low sensitivity (32.63%–57.69%) for lymph node metastasis.
• The federated-learning model achieved a sensitivity of 0.9231 without image data.
• Integrating MRI data further increased the federated-learning model's sensitivity to 0.9412.

Conclusions:

• Federated learning, particularly with MRI data integration, shows significant potential to revolutionize lymph node metastasis detection in gynecological cancers.
• This approach can enhance diagnostic accuracy, leading to more precise patient care and improved outcomes.
• The study highlights the transformative impact of AI and multimodal data in gynecological oncology diagnostics.