The Challenge of Ovarian Cancer
Ovarian cancer is often described as “rare, underfunded, and deadly,” according to Audra Moran, the head of the Ovarian Cancer Research Alliance (OCRA), a global charity based in New York. Like most cancers, catching it early can significantly improve outcomes.
This type of cancer typically begins in the fallopian tubes. By the time it reaches the ovaries, it might have already spread to other parts of the body. “You’d need to detect ovarian cancer at least five years before symptoms appear to have a real impact on survival,” Moran explained.
AI Offers New Hope
Advancements in artificial intelligence (AI) are offering hope. New blood tests powered by AI are being developed to detect ovarian cancer at its earliest stages. AI is also being used to speed up tests for other life-threatening infections, such as pneumonia.
Challenges in Training AI
Dr. Daniel Heller, a biomedical engineer at Memorial Sloan Kettering Cancer Center in New York, highlighted a key challenge in this field: ovarian cancer is rare, so there isn’t much data available for training AI algorithms. Even when data exists, it’s often kept in hospital systems with limited sharing between researchers.
Despite these challenges, the initial AI tests have shown promise. Dr. Heller called training the algorithm with data from only a few hundred patients a “Hail Mary pass,” but the system still managed to outperform current cancer biomarkers. As he explained, the AI’s accuracy is expected to improve as it is tested with larger datasets, much like self-driving cars get better with more road experience.
A Tool for Gynecological Diagnosis
Heller’s vision for this technology is ambitious. He hopes it can eventually help doctors quickly determine if a patient’s symptoms are more likely related to cancer or another gynecological condition. He estimates this tool could be available in about three to five years.
Breakthrough with Nanotube Technology
One breakthrough in his team’s research involves nanotubes—tiny carbon structures 50,000 times smaller than a human hair. These nanotubes emit fluorescent light when exposed to certain molecules in the blood, creating a unique signal. AI analyzes these signals to identify patterns too subtle for humans to detect.
Decoding these signals involves teaching AI algorithms which patterns are associated with ovarian cancer and which aren’t. Samples include blood from patients with other cancers or gynecological diseases to ensure accurate results. BBC has reported that this innovative approach could transform early cancer detection and save lives.
AI’s Impact Beyond Cancer
AI’s potential isn’t limited to cancer. For example, pneumonia can be deadly for cancer patients, especially since there are around 600 pathogens that can cause it. Doctors often need to run numerous tests to identify the specific infection, but AI is speeding up this process.
Simplifying Pneumonia Diagnosis
Karius, a California-based company, uses AI to pinpoint pneumonia pathogens within 24 hours and recommend the right antibiotics. “Before our test, diagnosing pneumonia involved 15 to 20 different tests during a patient’s first week in the hospital, costing about $20,000,” said Alec Ford, Karius’s CEO. Their AI compares patient samples to a vast database of microbial DNA, making this level of precision possible.
Unlocking Complex Patterns with AI
Researchers like Dr. Slavé Petrovski at AstraZeneca are also leveraging AI. Over the past two years, Petrovski’s team has developed a platform called Milton, which uses biomarkers to diagnose 120 diseases with over 90% accuracy. “These patterns are complex, involving multiple factors rather than a single biomarker,” Petrovski explained. BBC has noted that such breakthroughs are achievable only with AI’s ability to process massive datasets.
Data Sharing: A Key to Success
For ovarian cancer research, limited data remains a challenge. Moran emphasized that data sharing is crucial but often lacking. To address this, OCRA is creating a large-scale patient registry with electronic medical records to help train AI algorithms.
The Future of AI in Medicine
“It’s still early days for AI in medicine,” Moran added. “We’re in the wild west phase, but the potential is enormous.”
