corneal implants. (Photo provided by Kyungpook National University ICT Convergence Research Center)” data-type=”article”/>
Researchers (from the top left, Professors Daewoon Seong, Euimin Lee, Mansik Jeon, Professor Honggyun Kim, and Professor Jihyeon Kim) explain a deep learning-based model that automatically measures the thickness and volume of corneal implants. (Photo provided by Kyungpook National University ICT Convergence Research Center)
The joint research team of Kyungpook National University Department of Electronic and Electrical Engineering, Professor Man-sik Jeon, Professor Ji-hyun Kim, and Professor Hong-gyun Kim of Kyungpook National University Hospital (Ophthalmology) automatically measured the thickness and volume of the implant, which is one of the biocompatibility evaluation indicators used in corneal surgery prognosis, through a deep learning model. developed a measurement technique.
This study was published in npj Digital Medicine, an international academic journal in the field of ‘Health care sciences and services’ and an online medical journal of Nature Research.
Determining the biocompatibility of corneal implants is important in evaluating the prognosis of corneal surgery. The existing method of analyzing corneal implants is ultrasonic pachymetry, which analyzes the thickness by directly contacting the cornea with an ultrasonic measuring device.
In this study, the research team used ‘Optical Coherence Tomography (OCT)’ technology. OCT is a technology that obtains tissue images using the optical interference effect and interferometry. It is a device that analyzes the light reflected and returned when laser light is irradiated to the tissue to image the single-layer structure and characteristics of the tissue.
This technology can provide a variety of tissue information without contact, surgical invasion, destruction, or staining, so it is used as a representative technology in ophthalmology and is also used in various medical and industrial inspection equipment fields. This research team combined OCT with artificial intelligence technology to develop a deep learning model that can perform automatic region segmentation, measurement, and analysis of implants implanted in the cornea.
Among the developed deep learning models for corneal surgery prognosis evaluation, the ‘automatic region segmentation stage’ automatically separates and classifies the regions of the cornea and the artificial cornea in the OCT image of the cornea where the artificial cornea has been transplanted into a living body. At this stage, Kyungpook National University Eye Hospital was in charge of creating the dataset for transplant surgery and learning.
Afterwards, in the ‘measurement/analysis stage’, the resulting results were reconstructed in three dimensions to analyze the distribution of the remaining amount and thickness of the entire area of the corneal implant, and the cornea was not affected by the characteristics of each individual or the time elapsed after transplantation. It was confirmed that the implant could be analyzed.
This study was supported by the Ministry of Science and ICT and the National IT Planning and Evaluation Institute’s Regional Intelligence Innovation Talent Training Project and the National Research Foundation of Korea’s Mid-career Researcher Support Project. Professor Man-sik Jeon and Professor Ji-hyun Kim of the Department of Electronic Engineering at Kyungpook National University, and doctoral students Dae-un Seong and Ui-min Lee of the Department of Electronic and Electrical Engineering. , was conducted by a joint research team led by Professor Kim Yun-seok and Professor Kim Hong-gyun of the Department of Ophthalmology at Kyungpook National University Hospital.
Intern reporter Kim Nahye [email protected]
Great to have you here, Guest 1 and Guest 2. Let’s dive into the exciting topic of automated measurement of corneal implants using deep learning models. Guest 1, could you please tell us more about the significance of this research in the field of ophthalmology and corneal surgery?
Guest 1: Absolutely. This research has the potential to revolutionize corneal surgery by providing a non-invasive and accurate way to measure the biocompatibility of corneal implants. Current methods of analyzing implants involve physical contact, which can be uncomfortable for patients and may not always provide accurate results. With the development of this deep learning model, we can now use optical coherence tomography (OCT) technology to image the cornea and implants without contact, allowing for a more accurate assessment of thickness and volume. This information is crucial for evaluating the prognosis of corneal surgery, as it helps predict the success rate and potential complications.
Guest 2, could you explain how the deep learning model works? What were some of the challenges in developing such a model?
Guest 2: Sure, the deep learning model automatically segments and classifies regions of the cornea, the artificial cornea, and the remaining area of the implant in OCT images. This is a complex task as the boundaries between these regions can be quite undefined in many cases. We used a dataset created at Kyungpook National University Eye Hospital to train the model, and the model performs excellently even on unseen images. One of the major challenges was ensuring that the model could accurately distinguish between different regions and account for variations in the appearance of implants due to individual patient characteristics or the time since transplantation. We overcame this by using a three-dimensional reconstruction method to analyze the distribution of residual amounts and thickness of the entire implanted area.
Guest 1, how does the use of OCT technology in this research benefit patients undergoing corneal surgery?
Guest 1: OCT technology has been widely used in ophthalmology due to its ability to provide non-invasive imaging of tissues without contact or destruction. This study applies OCT to the specific context of corneal implant analysis,
