Estimation of Yttrium-90 Distribution in Liver Radioembolization using Computational Fluid Dynamics and Deep Neural...

Area of Science:

• Medical Physics
• Radiology
• Computational Fluid Dynamics

Background:

• Yttrium-90 (90Y) radioembolization is a liver cancer treatment using 90Y microspheres.
• Current dosimetry methods for prescribing 90Y activity lack accuracy, impacting treatment efficacy.
• Computational Fluid Dynamics (CFD) simulations offer improved dosimetry (CFDose) but are computationally expensive.

Purpose of the Study:

• To develop a faster dosimetry method for 90Y radioembolization.
• To accelerate the CFDose method for real-time treatment planning.
• To introduce a convolutional neural network (CNN) model for predicting 90Y distribution.

Main Methods:

• A CNN model was trained using CFD simulation results from a hepatocellular carcinoma patient.
• The model predicted 90Y distribution under varying downstream vasculature resistance.
• Performance was evaluated using mean squared error and prediction accuracy.

Main Results:

• The CNN model achieved high prediction accuracy, with an average difference of less than 1% between actual and predicted data.
• The model significantly reduced computation time compared to traditional CFD simulations.
• Accurate prediction of 90Y distribution was demonstrated.

Conclusions:

• The developed CNN model offers a significantly faster and accurate alternative for 90Y dosimetry in radioembolization.
• This AI-driven approach has the potential to enhance real-time treatment planning for liver cancer.
• Accelerated dosimetry can improve the effectiveness of Yttrium-90 radioembolization therapy.