Our Technology
How Cutisight Works
Pattern Recognition
Like learning to spot shapes in clouds and stars, Cutisight™ looks for patterns in skin images. During training, each pixel in an image is given a weightage, and linked to the described lesion and diagnosis. On passing thousands of images through the model repeatedly, the model learns what is common in all the images with a similar tag and uses this to analyse new data.
How It Sounds Human
It pieces together text from medical language it's seen before but it doesn't understand like a human. Think of it as a very advanced guesser!
Why It's Unique
It was trained after thorough understanding of why image models have failed in medicine. Each category is a diagnostic class, and there are 2000+ skin diseases. There is lack of data to make a pan-dermatology expert model. To overcome this challenge, Cutisight was trained on vast amounts of carefully curated data to give descriptions by identifying the primary and secondary lesions, colour, body site, and generate simple descriptions. Therefore it focuses on diagnosing the disease through morphology, and distribution not as a rote learned diagnosis.
Important to Remember
This is an under development model and can give confidently wrong answers. Always consult a qualified healthcare professional for diagnosis and treatment. Use appropriate medical literature for guidelines and protocols. This is not a replacement for a doctor.