研究目的
To develop a method for objectively evaluating retinal circulation by quantifying circulation-related parameters from fluorescein angiograms to aid in diagnosing retinal diseases.
研究成果
The proposed method successfully generates parametric images that quantify retinal blood flow and mean transit time, aiding in the detection of capillary nonperfusion areas. It reduces interobserver variability and standardizes FAG interpretation, with potential applications in telemedicine. Future improvements could address illumination issues and automate control point selection.
研究不足
The method may falsely indicate nonperfusion near the fundus edge due to uneven illumination in FAG images. Manual selection of control points can introduce errors, and the deconvolution process is sensitive to noise, with accuracy dependent on the choice of truncation threshold in TSVD. Illumination correction methods tested were ineffective or detrimental.
1：Experimental Design and Method Selection:
The study involves semiautomatic preprocessing and registration of FAG images, estimation of arterial input function using gamma-variate fitting, and computation of perfusion parameters (relative blood flow and mean transit time) via deconvolution based on truncated singular value decomposition (TSVD). Parametric color images are generated to visualize the results.
2：Sample Selection and Data Sources:
FAG images were acquired from an eye with symptoms of decreased vision or visual-field disturbance and evidence of retinal vein occlusion using a fundus camera. Digital images were taken at specific intervals post-injection of sodium fluorescein.
3：List of Experimental Equipment and Materials:
Fundus camera (Canon CF-60UVi; Canon Inc., Tokyo, Japan), sodium fluorescein dye, normal saline, computer for image processing.
4：Experimental Procedures and Operational Workflow:
Images were preprocessed to remove noise and text, registered to a reference image using semi-automatic methods with manual control point selection, arterial input function was estimated from a manually defined ROI, and parameters were computed using TSVD deconvolution.
5：Data Analysis Methods:
Gamma-variate fitting for curve modeling, Levenberg-Marquardt algorithm for nonlinear regression, TSVD for deconvolution to handle ill-posed problems, and generation of parametric images for visualization.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
