MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202641084340 A) filed by Dr. J. Anitha Ruth; M. Saraniya; G. Inbasakaran; and Dr. P. Visalakshi on July 09, 2026, for Automated Embryo Quality Assessment Using Otsu Segmentation And Embryonet Vgg16 Deep Learning Framework.
Inventors include Dr. J. Anitha Ruth; M. Saraniya; G. Inbasakaran; and Dr. P. Visalakshi.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: Infertility is the universal difficulties were the couples are not being able to give live birth occurrence. It is the problem which affects both men and women from various factors like hormonal imbalances, genetic problem, low sperm count, blocked fallopian tubes, ovulation disorders,-age^related“decline-and lifestyle_faeto,rs._:7Assisle_d Reproductive Technology or In-Vitro Fertilization encompasses medical procedures that offering options to individuals or couples to get pregnancy. The proposed study introduces a framework designed to automate the classification of embryos using time-lapse microscopic image datasets with a subset of 42 original embryo images which is augmented to 200 images of embryo both good and poor depending on the image readability and appropriateness. The first contribution of the research framework is the Otsu Segmentation which is the advanced pre-processing technique is used effectively to distinguish the embryo region between the foreground and background that act as the input for the model to obtain the best result. The second contribution was the EmbryoNic VGG 16 framework is employed for grading and highlighting the regions according to the cell development in embryo images which reduce the interference of the background, the result of the model has total accuracy of 88.1 %, the precision was 0.90, recall was 0.86 and the Fl- score was 0.88 respectively. Relative to the standard machine learning model, the proposed deep learning model is far more effective than KNN was 77.2 percent, Decision Tree was 82.3 percent. Random forest was 83.4 percent and the proposed EmbryoNet VGG 16 was achieved the highest result. The third contribution of the research was a Dual - Branch Convolutional Neural Network was implemented to enhance the quality evaluation of embryos by incorporating spatial image elements with morphological features that have clinical significance and the model result of precision was 0.85, recall was 0.90, Fl - score was 0.87 and the overall accuracy of the proposed model is the highest of 94.3 percent when we compared with the other model like CNN was 76.3 percent, ResNet 50 was 80.8 percent, MobileNet V2 was 82 percent. To summarize, this study shows that proposed designed deep learning model can have the outstanding prediction of embryo which highlights its effectiveness in automated evaluation, the finding shows how well the system improves the reliability and efficiency of assessing the best quality of embryo for the successfuL implantation
Disclaimer: Curated by HT Syndication.
