Computer Diagnoses Cataracts As Well As Eye Doctors Can

Congenital cataracts is a rare eye illness. It is characterized by clouding of the eye lenses in babies that causes vision to be unclear and obstructed. It is one of the chief causes of blindness in children and is easy to treat. However, treatment for the condition is extremely expensive and one of the most difficult types of interventions in the field of ophthalmology.

According to Professor Lin Haotian of the Sun Yat-Sen University, who is one of the authors of the study, around 200,000 children suffer from congenital cataracts and are blind in both eyes as a result. He stated that many others are affected by partial cataracts, but the disease tends to progress as the children age, leading to even more difficulty with their vision.

Dr. Lin and his team wanted to make it easier to diagnose congenital cataracts in more locations around the world. As a result, they developed a convolutional neural network based artificial intelligence platform called CC-Cruiser. The platform is able to identify, evaluate and recommend treatment for the eye condition.

Together, the researchers created the AI based on a database under the Childhood Cataract Program of the Chinese Ministry of Health. They equipped CC-Cruiser with a set of 476 photos of normal eyes and 410 of eyes afflicted by congenital cataracts. Each set of images was labeled by a pair of knowledgeable ophthalmologists.

Afterward, the researchers tested out CC-Cruiser in a number of complex real-world scenarios. One of those was within a clinical trial in multiple hospitals. There were other website based tests and one test that compared the performance of CC-Cruiser versus human ophthalmologists.

Lin reported to Asian Scientist Magazine that the overall accuracy of CC-Cruiser’s performance was extremely similar to that of an experienced ophthalmologist. He added that the AI was successful at diagnosing all of the potential patients within 50 cases and that, in comparison, the human doctors were incorrect with their diagnoses of some cases.

Lin concluded that CC-Cruiser delivered accurate suggestions regarding treatment for each patient who was found to require surgery, “with five false positives.”

In addition, Lin acknowledged that CC-Cruiser’s performance could be improved even more by giving it a larger dataset. As a result, the researchers created a collaborative cloud platform so that data could be integrated and CC-Cruiser’s functionality can improve even more by complementing current approaches at addressing rare illnesses.

Although the results CC-Cruiser has garnered are promising, Lin believes it would be used to augment and not actually replace doctors. He stressed that machines may have precision, but human abilities such as interaction and communication are indispensable in the medical field.