6 results about "Retina Neovascularization" patented technology

The invention discloses a retina neovascularization segmentation method based on a dual-path multi-scale attention network, and belongs to the technical field of medical information processing, and the method comprises the steps: determining a WF-OCTA image of a retina neovascularization, drawing a segmentation mask on the WF-OCTA image, and obtaining input data; multi-scale features of input data are extracted through an encoder-decoder, jump connection is established, and multi-stage input feature tensors are obtained; performing two-way multi-scale feature mining on the input feature tensor, capturing form information and context information in parallel, and obtaining a residual enhanced output feature tensor; statistical space-spectrum attention enhancement is carried out on the residual error enhancement output feature tensor, channel re-calibration and statistical magnitude-based space fine positioning are carried out, and a statistical enhancement output feature tensor is obtained; and carrying out fusion statistics on the output feature tensor in the decoding path, carrying out step-by-step reconstruction, obtaining a pixel-level probability graph, carrying out training by using mixed loss, and outputting a segmentation result.

Microparticulate (MP) formulations formed from one or more poly(hydroxyacid) polymers having a molecular weight ranging from 5kD to 60kD, and one or more active agents are injected into the eye of a subject to address eye disorders. The MP formulations assure high drug loading and extended delivery of the active agent of six to twelve months. The active agents may include peroxisome proliferator-activated receptor alpha (PPARα) signaling agonists such as PPARα agonist A190 (IUPAC name 3-((4-((4-fluorobenzyl)oxy)- 3- methylbenzyl)amino)benzoic acid). The formulations have therapeutic and protective effects against retinal degeneration diseases such as age-related macular degeneration (AMD), but may also be used for treating or relieving the symptoms of retinal inflammation, retinal neovascularization, retinal vascular leakage, retinopathy of prematurity (ROP), diabetic retinopathy (DR), and diabetic macular edema (DME).

The application provides an animal model of high-oxygen-induced retinal neovascular disease and a preparation method thereof. The method comprises the following steps: placing a mouse born for 3-7 days in an environment with an oxygen concentration of 75-85%, and maintaining for 10-30 days. The method can successfully induce a long-acting retinal neovascular disease model, has a high modeling rate, fewer complications, more persistent retinal neovascular lesions, and a model duration of 55 days. The animal model provides a more optimal model for exploring long-acting anti-retinal neovascular related drugs.

The invention relates to an application of leucine-rich transmembrane fibronectin 2 (FLRT2) in preparation of a medicine for intervening abnormal retinal neovascularization. The research finds that the leucine-rich transmembrane fibronectin 2 can promote the generation of retinal neovascularization, and the generation of pathological abnormal retinal neovascularization can be inhibited by inhibiting the expression of the leucine-rich transmembrane fibronectin 2; the leucine-rich transmembrane fibronectin 2 can be used as a new target for preventing and / or intervening eye retinal neovascularization related diseases, and provides a new direction and strategy for intervention of retinal abnormal neovascularization related diseases. The leucine-rich transmembrane fibronectin 2 inhibitor is any one or any combination of small interfering ribonucleic acid selected from small interfering ribonucleic acid 1, small interfering ribonucleic acid 2, small interfering ribonucleic acid 3, small interfering ribonucleic acid 4, small interfering ribonucleic acid 5 and small interfering ribonucleic acid 6.

The invention discloses a retina neovascularization automatic typing method based on a scale perception network, and belongs to the field of medical image intelligent analysis and computer-aided diagnosis, and the method comprises the steps: inputting an OCTA image of a retinopathy patient in a proliferation period, inputting the OCTA image into a pre-trained scale perception typing model, carrying out image standardization preprocessing on the OCTA image by a self-adaptive preprocessing module in the scale perception parting model to obtain a standardized image; the deep topological feature extraction backbone network performs multi-scale blood vessel morphological feature extraction based on the standardized image; and the prediction output module outputs new blood vessel form typing results according to the extracted multi-scale blood vessel form characteristics, and the typing results comprise a single trunk branch type, a multi-source mesh type and an anastomosis ring type. According to the method, rapid, accurate and objective automatic typing of the new blood vessel in the OCTA image is realized, and auxiliary decision support is provided for individualized accurate treatment of proliferation phase retinopathy.

The application discloses a retinal neovascularization segmentation method based on a double-path multi-scale attention network, and belongs to the technical field of medical information processing. The method comprises the following steps: determining a WF-OCTA image of retinal neovascularization, drawing a segmentation mask on the WF-OCTA image, and obtaining input data; extracting multi-scale features of the input data through an encoder-decoder and establishing a skip connection to obtain a multi-level input feature tensor; performing double-path multi-scale feature mining on the input feature tensor, capturing morphological information and context information in parallel, and obtaining a residual enhanced output feature tensor; performing statistical spatial-spectral attention enhancement on the residual enhanced output feature tensor, performing channel re-labeling and spatial fine positioning based on a statistical quantity, and obtaining a statistical enhanced output feature tensor; fusing the statistical enhanced output feature tensor in a decoding path and gradually reconstructing to obtain a pixel-level probability map, training the pixel-level probability map through a hybrid loss, and outputting a segmentation result.