Deep Residual Inception Encoder-Decoder Network for Medical Imaging Synthesis.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Machine Learning

Background:

• Image synthesis is crucial in precision medicine for unavailable medical imaging.
• Convolutional Neural Networks (CNNs) are effective for image synthesis due to their complex architectures.

Purpose of the Study:

• To propose a novel Residual Inception Encoder-Decoder Network (RIED-Net) for medical image synthesis.
• To evaluate RIED-Net's performance against existing models.

Main Methods:

• Developed a new RIED-Net architecture for nonlinear image mapping.
• Compared RIED-Net with synthetic CT deep convolutional neural network (sCT-DCNN) and shallow CNN.
• Utilized mammogram and neuroimaging datasets for evaluation.

Main Results:

• RIED-Net demonstrated superior performance over sCT-DCNN and shallow CNN.
• The proposed model achieved significant improvements in structural similarity index, mean absolute percent error, and peak signal-to-noise ratio.

Conclusions:

• RIED-Net is a highly effective architecture for medical image synthesis.
• The proposed method offers a promising solution for generating high-fidelity medical images in precision medicine.