Retinopathy of prematurity (ROP) is the leading cause of childhood blindness worldwide. Automated ROP detection system is always desired to improve the ROP care. Professor Yi Zhang and his colleague at the Machine Intelligence Laboratory of Sichuan University, Sichuan Academy of Medical Sciences and Sichuan Provincial People’s Hospital, and Chengdu Women & Children’s Central Hospital started the project of ROP Detection by using AI in the July 6th, 2017. Recently, they found that deep neural networks (DNNs) perform well in the detection of ROP in retinal fundus images. They build a large scale ROP detection dataset by labeling the retinal fundus images by adequately trained clinical ophthalmologist and developed two DNN models for ROP identification and grading, respectively. Furthermore, they developed an automated ROP detection system called DeepROP by employing the proposed DNN models. It turned out that the proposed DNN models obtained high sensitivity and specificity and outperformed some human experts. Their results were published in EBioMedicine, whose two leading brands are the Lancet and Cell Press.
“DNN is a class of biological inspired computational model that is an important method towards artificial intelligence. It could learn the abstract features from big data and have achieved many impressive results in practical applications, including medical big data analysis”, said professor Yi Zhang. It is known that the medical big data research group at Machine Intelligence Laboratory has already developed many AI systems to aid the diagnose of diseases, including DR, breast cancer, and lung node.
Access the EBioMedicine study, “Automated retinopathy of prematurity screening using Deep Neural Networks”, EbioMedicine, 2018. DOI:10.1016/j.ebiom.2018.08.033
For more information on the research and application of medical big data analysis by using DNNs, go to http://zyimed.machineilab.org