Automated unsupervised recognition of cone photoreceptor cells in adaptive optics scanning laser ophthalmoscope images

Physics

Identifying the specific type of cell, known as the cone photoreceptor cell, is essential for accurately diagnosing and treating many eye diseases. This research presents a novel automated approach that use unsupervised learning to detect CPCs in images obtained from adaptive optics scanning laser ophthalmoscopes. This approach is founded on the fundamental concepts of machine learning. The steps involved in this approach include of estimating CPC numbers, rectifying bias fields, autonomously recognizing CPCs, and integrating data from nearby sites during CPC identification. This procedure is done sequentially. The results of our study demonstrated that the proposed method surpassed the manually created techniques in terms of recall (84.4%), accuracy (92.9%), and F1 score (88.4%). Based on the results of this comparison, the proposed approach demonstrated satisfactory performance. Our approach is capable of processing AO-SLO images of both normal and diseased retinas, including those with different CPC densities, as well as images of diabetic retinopathy. The findings demonstrated the high precision of our approach in identifying circulating progenitor cells (CPCs) in eye tissue samples from both healthy individuals and those with diabetic retinopathy.The objective of this project is to create a completely automated and unsupervised method for identifying cone photoreceptor cells in AOSLO (Adaptive Optics Scanning Laser Ophthalmoscope) images. An exhaustive examination of cone photoreceptors can offer valuable understanding into numerous retinal illnesses; these cells are essential for vision. These cells are famously challenging to detect using standard approaches due to their high degree of physical interaction and effort required. The objective of this project is to develop an innovative approach that utilizes cutting-edge machine learning algorithms to detect and classify cone cells in AOSLO images, without requiring pre-existing labeled training data. The program employs image processing techniques and feature extraction algorithms to differentiate cone cells based on their distinct structural attributes. The objective of this study is to mechanize the process of identifying retinal images in order to enhance efficiency and precision. This has the capacity to improve both the diagnostic capacities and the monitoring of therapy for retinal illnesses.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.