Artificial intelligence expert enhances cancer diagnostic testing & delivers value to your NGS Data!
Find out about our novel approach to cancer diagnosis. We’ll present our very high accuracy and show how many lives can be saved if cancer is detected early. Although not covered in this article, our platform can help develop other analyses as well, e.g., AI platform for NGS data analysis.
Every year, cancer claims the lives of over 8 million people. To fight this epidemic, the World Health Organization is encouraging governments to focus on early, non-invasive detection, which has been shown to dramatically increase treatment success. Unfortunately, current diagnostic testing and screening tools lack the necessary accuracy rates, meaning many patients remain undiagnosed.
To enable early diagnoses and screening for a wide array of cancer types, Romanian start-up company Artificial Intelligence Expert (AIE), with support from the EU, developed an extremely accurate diagnostic tool. Called AI-MICADIS, the solution uses circulating microRNAs instead of the more commonly used circulating tumour DNA and circulating cancer cells. It also uses innovative deep learning artificial intelligence (AI) algorithms to provide real-time analysis of polymerase chain reaction (PCR), microarray, and next generation sequencing (NGS) data – all essential factors in detecting the earliest stages of cancer progression.
“By selecting informative molecular alterations and using artificial intelligence, we developed the best non-invasive multi-cancer diagnosis and early detection tests available,” says Alexandru Floares, an AIE researcher and coordinator of the EU-funded AI-MICADIS project.
During the AI-MICADIS project, AIE conducted a feasibility study to test the tool, identify potential partners, set out a regulatory path, and optimise the best strategy for bringing AI-MICADIS to market.
Near perfect accuracy
As to the feasibility of the AI-MICADIS tool itself, the project tested it on 6 000 patients. From this work, researchers were able to extend its use from nine to thirteen cancer types. They also added new AI methods capable of learning from only a few cases – an advancement that enables faster and less costly validation of the tests.
Originally, the AI-MICADIS tool was configured to discriminate between cancer and normal cells. But after testing the tool with oncologists, researchers learned there was more interest in discriminating between cancer and benign diseases. Unfortunately, this capability was not possible with the system’s current molecular set-up. In fact, when it came to distinguishing the malignant from the benign, AI only achieved 67 % accuracy – a far cry from the near perfect diagnosis rates the team was hoping for.
Instead of throwing in the towel, the AIE team went back to the drawing board to reframe the problem and develop new techniques. “In the end, we proved a remarkable 99 % accuracy rate, along with the tool’s ability to discriminate between malignant and benign diseases,” explains Floares. “Furthermore, we demonstrated that the use of data science and AI to integrate multiple studies outperformed the use of a single extensive study in terms of costs, efficiency and results.”
A solid business case
As to the business case for AI-MICADIS, Floares notes that the project allowed his team to learn more about their potential customers and to better understand the complex regulatory environments of both the EU and United States. “We made our team stronger on the business side and we identified potential partners. We also won other project competitions and identified potential investors, which puts us in a good position to move towards marketisation,” Floares says.
AIE is currently working to secure the funding needed to bring the AI-MICADIS tool to the global market, while also adding new cancer types to the system’s portfolio.
AI-MICADIS, artificial intelligence, AI, cancer, World Health Organization, WHO, diagnostic testing, deep learning
AI-MICADIS is using AI and Digital Automation to create the most performant non-invasive molecular and imaging intelligent systems for early diagnosis and personalised treatment of patients, e.g., we are making early cancer diagnosis a routine.