A Helping Hand

In an exciting follow up to “AI to Interpret,” published in The Ophthalmologist back in 2020, the first artificial intelligence (AI) platform for OCT scan analysis has now launched online for public access with FDA clearance currently in process (1). The new Altris AI platform will help alleviate the compounded pressures of the fight against preventable blindness – affecting more than one billion people worldwide – and the global shortage of optometrists (2,3).

Able to detect more than 100 retina pathologies and pathological signs – both widespread and rare – and differentiate between pathological and non-pathological OCT scans within minutes, the new platform provides a promising decision making support tool to help ophthalmologists in their daily practice. Eye care specialists from 140 countries have already registered onto the platform and tested its benefits in clinical practice.

How was the AI model created? First, the Altris AI medical team collected and labeled thousands of OCT scans under my supervision. Our team of ten experts then marked pathologies and pathological signs on all OCT scans inside a special program, preparing them for the model. This activity was followed by the implementation and training of the AI model. Popular neural network architectures for medical imaging, such as UNet and FPn, were used as a starting point alongside necessary proprietary changes that ensured the model was both adaptable yet capable of handling the complex data involved.

This AI tool – along with many others –  is here to help clinicians, offering a second opinion on OCT scans and automating some of the most repetitive tasks of ophthalmology. More than this, the combined knowledge of eye care specialists and AI models may help to reduce human errors among less experienced eye care specialists.