Potential for Artificial Intelligence in Medical Imaging and Diagnostics Demonstrated by Johns Hopkins Researchers

Potential for Artificial Intelligence in Medical Imaging and Diagnostics Demonstrated by Johns Hopkins Researchers

Age-related macular degeneration (AMD) is the leading cause of blindness for the over 50s. The condition causes lesions to form that blur the central part of vision. The condition is progressive and can seriously affect quality of life, making driving, reading and even recognizing faces difficult.

As the condition progresses, the central part of the field of vision may be completely lost. Early diagnosis is essential to ensure treatment is provided to prevent the condition from getting worse. Once vision loss has occurred it is usually irreversible.

Researchers at Johns Hopkins University Applied Physics Laboratory in Maryland have been working on an AI system that can help ophthalmologists diagnose the condition more quickly to ensure treatment can be started before irreversible vision loss has occurred. The researchers have been able to train their AI system using the digital image library of the Johns Hopkins Wilmer Eye Institute.

The team has achieved great success. The AI system has been trained and can now diagnose the condition with the same level of accuracy as highly trained ophthalmologists. There is tremendous potential for the system to be used to automatically grade images to speed up the diagnosis of the condition as well be used for monitoring individuals with the early stages of the disease.

The researchers have been working with the Johns Hopkins Wilmer Eye Institute since 2015 and have recently started using the system to help with the diagnosis of other eye conditions such as diabetic retinopathy, training the system to characterize retinal layers in optical coherence tomography. The system could also be used to characterize vascular and neurodegenerative pathologies.

The huge quantities of data that can be read by the AI system means that in addition to characterizing images, it would be possible to incorporate many other types of information to help determine the risk of someone developing AMD – information such as the level of sunlight exposure or smoking, which can increase the likelihood of someone developing the disease.

The researchers have recently started collaborating with several other eye research centers, including the Singapore National Eye Center, to train the AI system further, specifically on Asian ethnic groups. Further collaborations are planned with eye institutes in Brazil, France, and Thailand to determine how the system can be adapted to work with different ethnic groups and for other eye conditions.

The researchers have detailed the progress made so far with their AI system its requirements, and future work that is necessary before the system can be used for clinical and point of care scenarios. The article – AI for Medical Imaging Goes Deep – has been published in Nature Medicine, volume 24, pages 539–540 (2018).

Leave a Reply

Your email address will not be published. Required fields are marked *