Top image: Glenn Linde and Dr Hong Sheng Chiong.
MedicMind is a world-first artificial intelligence (AI) medical platform for medical researchers and clinicians to create AI to auto-diagnose a large range of diseases based on a single photograph. oDocs, which created the system for MedicMind, says the AI system is working well, and is set for beta testing.
MedicMind spokesperson Glenn Linde says there’s no other system that is specifically for medical researchers. “The closest thing to this system is IBM’s Watson, but it is not designed specifically for medical researchers,” he says. “With MedicMind a smartphone app could use a neural network created with MedicMind trained to detect diseases like glaucoma, melanoma, and diabetic eye disease.
“A lay person using a smartphone app could detect melanoma themselves or a clinician in a hospital could detect eye degeneration from an image of a patient’s retina.
“We have a trainee eye doctor, Clark Stevenson, who plans to use neural networks generated with MedicMind on some patients or some of his students in the future. Stevenson is already testing with Dunedin hospital students.”
MedicMind is a cloud-based platform where medical researchers can upload images for their own AI training. In other words, if they want to create a medical neural network that can detect glaucoma, they can do so with no prior knowledge of computer coding or programming. Says Linde: “Researchers could take the neural network created on MedicMind and put it on a smartphone or PC or whatever device they choose, and that device will be able to detect glaucoma from a photo.”
MedicMind founder Dr Hong Sheng Chiong, an ophthalmologist registrar at Dunedin hospital, wanted to use AI with image classification for improving medical diagnosis for patients. He also wanted to use image classification on thousands of 3D scans of people’s heads he scanned in the famous Dunedin Study, and predict things like anaemia just from faces.
Dr Hong says he envisioned a system that would enable researchers and clinicians to develop their own AI without prior knowledge of coding or programming. “There are many people like myself out there with a vision to develop the next generation AI for improving healthcare, but the technical skills and cost are overwhelming,” he says. “Research in the field of medical AI is not accessible to the general public at all.”
Listen to this podcast with Amazon's Alayna Van Dervort on using AI for good:
Linde says he initially thought it unrealistic to build an AI system with a simple drag and drop interface because of the complexity, but realised that with cloud technology there were unlimited resources for AI development that could be utilised.
MedicMind is built with a similar AI framework as Google’s DeepMind, using a state-of-the-art neural network model. Since it’s on the cloud, no special hardware is needed for users to access it; all that’s needed is a computer and access to internet.
Dr Hong and Linde say they built the system to improve diagnostic healthcare and make AI more accessible. If medical professionals start using it to improve patients’ lives, then it’s mission accomplished. The true definition of using AI for good if there ever was one.
Hear this podcast with Touch Tech CEO Andrew Smith about SAM, the artificial intelligence-powered virtual politician that may one day actually run for office:
Idealog has been covering the most interesting people, businesses and issues from the fields of innovation, design, technology and urban development for over 12 years. And we're asking for your support so we can keep telling those stories, inspire more entrepreneurs to start their own businesses and keep pushing New Zealand forward. Give over $5 a month and you will not only be supporting New Zealand innovation, but you’ll also receive a print subscription and a copy of the new book by David Downs and Dr. Michelle Dickinson, No. 8 Recharged (while stocks last).
Idealog is part of ICG. We work with clients like Woolworths New Zealand, All Good, Huffer, Liquorland, Resene, Citta Design, TVNZ, Spark and FCB on their event activations, in-store, in-office or out-of-home signage, content creation and vehicle wraps.