Machine learning algorithms are programmed to detect potential red flags in blood test results.
This is the third installment in a six-part series examining how artificial intelligence is transforming medical research and therapies.
“Ovarian cancer is ‘uncommon, underfunded, and fatal,'” states Audra Moran, the president of the Ovarian Cancer Research Alliance (Ocra), a worldwide non-profit organization headquartered in New York City.
Similar to all forms of cancer, early detection leads to improved outcomes.
The majority of ovarian cancer cases originate in the fallopian tubes, and by the time it reaches the ovaries, the disease may have metastasized to other regions.
“To impact mortality rates, ovarian cancer detection might be required five years before any symptoms manifest,” remarks Ms Moran.
However, novel blood tests are surfacing that harness the capabilities of artificial intelligence (AI) to detect indications of cancer during its initial phases.
Moreover, AI is not limited to cancer detection, it can expedite other blood assessments for potentially fatal ailments such as pneumonia.
Dr. Daniel Heller is a biomedical engineer employed at the Memorial Sloan Kettering Cancer Center located in New York.
Dr. Heller's team has developed a testing technology that utilizes nanotubes – minute carbon tubes approximately 50,000 times thinner than a human hair's diameter.
Roughly two decades ago, researchers started identifying nanotubes capable of emitting fluorescent illumination.
Over the last ten years, scientists have learned to modify the properties of these nanotubes so they can detect nearly any substance present in blood.
Presently, it's feasible to introduce millions of nanotubes into a blood sample and have them emit varying light wavelengths based on the substances they bind to.
However, the challenge remained in decoding the signal, which Dr. Heller compared to identifying a fingerprint match.
Here, the fingerprint refers to the pattern of molecules adhering to sensors, with varying degrees of sensitivity and binding affinities.
However, the patterns are too intricate for a person to discern.
“We can examine the information and it will make no sense whatsoever,” he states. “We can solely identify the divergent patterns through AI.”
Deciphering the nanotube data involved inputting the information into a machine-learning algorithm, and informing the algorithm which samples originated from ovarian cancer patients, and which were from individuals without the disease.
These comprised blood samples from individuals with different cancer types, or other gynecological conditions that could potentially be mistaken for ovarian cancer.
A major hurdle in leveraging AI to create blood tests for ovarian cancer research is the relatively low prevalence of the disease, which restricts the data available for training algorithms.
Furthermore, a significant portion of that data is compartmentalized within hospitals that provided treatment, with limited data exchange for research purposes.
Dr. Heller characterizes training the algorithm on accessible data from merely a couple hundred patients as a “desperate attempt.”
However, he states the AI could achieve higher precision than the top cancer biomarkers currently accessible – and that was merely the initial attempt.
The system is undergoing additional research to determine if its performance can be enhanced by utilizing larger collections of sensors and samples from numerous more patients. Increased data can refine the algorithm, similar to how algorithms for autonomous vehicles can improve with more road testing.
Dr. Heller harbors substantial expectations for the technology.
“Our aspiration is to categorize all gynecological ailments – so that when a patient presents with symptoms, we can provide medical practitioners a rapid diagnostic aid to discern whether it's probably cancerous or pinpoint the specific type of cancer,” remarked the expert.
According to Dr Heller, this technological advancement could be realized within “three to five years”.
Artificial intelligence's potential applications extend beyond early detection, as it is also expediting the processing of various blood tests.
For cancer patients, contracting pneumonia can be life-threatening, and with approximately 600 distinct microorganisms capable of causing the illness, physicians must perform numerous tests to pinpoint the underlying infection.
However, novel varieties of blood tests are streamlining and accelerating the procedure.
Karuis, a California-based firm, utilizes artificial intelligence (AI) to pinpoint the precise pneumonia pathogen within 24 hours and recommend the appropriate antibiotic.
“Prior to our test, a pneumonia patient would undergo 15 to 20 various tests to diagnose their infection within just the first week of hospitalization – amounting to roughly $20,000 in testing costs,” remarked Alec Ford, CEO of Karius.
Karius possesses a database of microbial DNA comprising tens of billions of data points. Patient test samples can be matched against this database to pinpoint the precise pathogen.
According to Mr. Ford, such a feat would have been unattainable without the aid of Artificial Intelligence.
A hurdle arises as scientists may not fully comprehend the associations an AI could forge between biomarkers in tests and various ailments.
During the past couple of years, Dr. Slavé Petrovski has created an AI platform dubbed Milton that leverages biomarkers in UK biobank data to pinpoint 120 diseases with an accuracy exceeding 90%.
Discerning patterns within such a vast quantity of data is a capability exclusive to artificial intelligence.
“Such patterns are frequently intricate, where a single biomarker may not suffice, necessitating consideration of the entire picture,” remarks Dr. Petrovski, a researcher at the pharmaceutical behemoth AstraZeneca.
Dr. Heller employs a comparable pattern-matching methodology in his research on ovarian cancer.
“The sensor bonds with and reacts to proteins and tiny molecules present in the bloodstream, but we're unaware which specific proteins or molecules are indicative of cancer,” he states.
On a broader scale, data or the absence of it remains a disadvantage.
“Individuals are not divulging their information, or there lacks a system to facilitate it,” states Ms. Moran.
Ocra is financing an extensive patient database, encompassing electronic health records of individuals who've permitted researchers to train algorithms utilizing their data.
This News Story is brought to you by QLM Business News, your Digital Media Channel
For more business news stories also follow us on Facebook, X and Youtube.
Disclaimer: All images presented herein are intended solely for illustrative purposes and may not accurately depict the true likeness of the subjects, objects, or individuals referenced in the accompanying news stories.
