Now for a little over a decade, the value delivered by Artificial intelligence (AI) In Medical diagnosis Realizing its role in areas that have long posed serious challenges for healthcare professionals, it continues to grow. Big killers like Tuberculosis, Cancer and more recently Kovid are being tackled by the power of AI to deliver early detection (or in the case of Kovid affect the lungs), thus saving millions of lives worldwide.
AI models are now used for the diagnosis of preliminary examinations and MRI scans, medical, microscopic and ultrasonic imaging. Some common applications include tumor detection, its size, and measuring blood flow at the microscopic level. While sophisticated algorithms help AI systems learn from every ‘transaction’ or data they place, their effectiveness largely depends on the volume and quality of the data they are given. This is a significant challenge, primarily in healthcare or, more precisely, in diagnostics.
To work around this challenge, AI specialists rely on what is commonly known as data growth and integration techniques, a process by which large amounts of ‘synthetic data’ are created that behave like clones of real data. Intermix One such proprietary technology for generating high-quality artificial data that now has immense potential in healthcare.
To understand how Intermix works, we need to understand AI at its very basic level. AI models mimic the human brain (hence the name neural network). The human brain is one of the most complex systems that deals with a large number of parameters and tons of data. Since AI models sometimes need to make decisions for complex and challenging data such as tumor diagnosis, we need to show them lots of examples so that they can clearly understand the difference. It may not be easy for a human to distinguish between a benign and malignant tumor, unless it is done by an experienced physician who has thousands of such scans and decades of experience. Intermix certainly does that.
Intermix “overlays” two data points on top of each other. For images, think of this as choosing a level of transparency on both images and placing them on top of each other. Experimentally, the use of these synthetic specimens improves the accuracy regarding the prediction of models for speech and medical imaging problems of more than 5% in multiple cases on multiple issues and benchmark datasets. Also, since the image from the actual data is always mixed to generate synthetic data, there is no apparent risk of adding external noise or distortion.
Imagine 30 scans of malignant breast cancer tumors and 500 benign breast cancer tumors. By overlaying and blending these images into different ratio, contrast and transparency layers, we can create many new images that will be a combination of these images. For example, on a very weak 10% visible image of a benign tumor 90% malignant tumor image can be fed to an AI model that will learn this new distribution. Such multiple models increase the model’s confidence and learning accuracy.
The cost of Intermix, Indigenously owned technology to create artificial data, has tremendous potential in a country like India. As detailed here, its application in healthcare alone will be huge. Access to medical imaging and diagnosis in rural parts of the country will be significantly improved as it can replace the rare and expensive professionals and machines. It is too early to predict the social and economic value it may provide, but it will certainly be very significant.
It will also be a big addition to the field of voice augmentation to train virtual assistants like Apple’s Siri and Amazon’s Alexa. As things stand now, it is much easier to get voice data for widely spoken languages like Hindi and English. According to the latest census, Indians speak more than that 19,500 languages and dialects As their mother tongues. With Intermix, the search engine will be even more effective in rural languages. It will help build recommendation algorithms for Netflix, Instagram, ShareChat, etc. in regional languages and regional content and help create a whole new market for these services.
Not only that, but a solution like Intermix could be one of the most effective bridges across the digital divide across the country. Whether it is in AI-based healthcare or many other applications have a big impact in the country, data growth technologies like InterMix can be a potential game-changer for a country like India.
By Ramit SahniAI Instructor at DSML and Scalar
(Disclaimer: The opinions expressed are those of the author only and ETHealthworld is not required to subscribe to them.
