Project Details
Description
The mechanism of Near Infrared Diffuse Optical Imaging (NIR DOI) involves utilizing the differences in the tissue optical coefficients (such as absorption and scattering coefficients) of human tissues in a specific near-infrared wavelength range to distinguish tissue abnormalities, particularly in breast tumors. The development of related computational software and modular equipment for DOI imaging began in the mid-1990s. Image computation is based on the model (diffusion equation) and involves discrete forward calculations and iterative inverse calculations to obtain images of tissue absorption and scattering coefficients.Researchers in image reconstruction need to possess medical physics knowledge and numerical optimization algorithm capabilities. Even software users, in addition to establishing two-dimensional and three-dimensional computational models, must have the ability to adjust regularization computation parameters for quality assessment, which can be influenced by individual backgrounds and experience, affecting the quality of image computation.In the past decade, the proposal and widespread application of deep learning neural networks have led to the development of data-based computational models in the field of DOI tissue optical coefficient imaging technology. This involves the creation of models based on classification and regression neural networks using large datasets. Therefore, in the field of DOI tissue optical coefficient imaging technology, computational models based on data have started to emerge, utilizing extensive simulated data and partial experimental data, including peripheral light information (intensity and phase) of the subject under test and computed optical coefficient images.In the planned two-year project, the goal is to complete a diffusion light imaging computational platform based on deep learning models. In addition to the already implemented and validated batch normalization CNN model, light information image neural model, and the ongoing work on the periodic neural network model, the project proposes and implements research projects such as (i) hybrid neural network models for sensing/imaging, (ii) deep learning model ablation studies, and (iii) irregular profile shape transfer learning models. All these research projects will undergo numerical simulations and phantom experiments to facilitate subsequent human trials and clinical applications in breast surgery.
Status | Finished |
---|---|
Effective start/end date | 15/08/23 → 14/08/24 |
UN Sustainable Development Goals
In 2015, UN member states agreed to 17 global Sustainable Development Goals (SDGs) to end poverty, protect the planet and ensure prosperity for all. This project contributes towards the following SDG(s):
Keywords
- Diffuse Optical Tomography
- Breast Tumor
- Deep Learning Neural Network Model
- Transfer Learning Neural Network Model
- Model Ablation Study
Fingerprint
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.