Artificial intelligence has been making some big leaps in the medical world. A new study published in The Lancet Oncology shows that AI can be trained to detect and evaluate prostate cancer from needle biopsy samples. The AI system was on par with world-leading pathologists in its judgment on the severity of prostate tumors.
“Our results show that it is possible to train an AI system to detect and grade prostate cancer on the same level as leading experts,” said Dr. Martin Eklund, the lead study author. Eklund is the associate professor in the Department of Medical Epidemiology and Biostatistics at Karolinska Institutet in Sweden.
“This has the potential to significantly reduce the workload of uropathologists and allow them to focus on the most difficult cases,” Eklund said.
Improving treatment and reducing medical expenses
Not only could the AI system reduce the workload of uropathologists but the system could also reduce medical expenses for cancer patients. About 38.4% of people worldwide will be diagnosed with cancer at some point during their lifetime and 42% of U.S. patients deplete their life savings in their first two years of treatment. The average expenditure for a single office-based physician visit is $155.
AI technology could help to reduce the number of appointments and screenings patients need to go through before they’re given a clear diagnosis. What’s more, they can receive cancer screening from a system that’s on par with some of the most experienced doctors.
“The idea is not that AI should replace the human involvement, but rather act as a safety net to ensure that pathologists don’t miss some cancers, and assist in the standardization of grading,” said Eklund. “It could also serve as an alternative in parts of the world where there is a complete lack of pathologic expertise today.”
One of the key issues in today’s prostate pathology is that different pathologists can reach different conclusions while assessing biopsies even when studying the same samples. This can lead to clinical issues where treatment options for patients are made based on ambiguous information.
This is a problem because cancer treatments can often cause other side effects including nausea, vomiting, fatigue, low white blood cell count, and lymphedema. Hair loss is also a common side effect of cancer treatment, which isn’t as serious as other side effects but can increase a patients’ feelings of distress.
Hair loss can be especially distressing for men, two-thirds of whom will experience some type of hair loss by age 35 even without cancer treatment. Men’s grooming in the U.S. alone is a $4 billion industry.
“Appearance-related concerns are normal and to be expected throughout the cancer experience,” said Carrie Panzer, a social worker at the Memorial Sloan Kettering Cancer Center. “Understanding this can be a key component in how you cope, now and even years into survivorship.”
The new AI system could help to reduce treatments based on subjective information and could also help to reduce the workload in prostate pathology. Over 1 million people undergo a prostate biopsy every year, producing over 10 million tissue samples that need to be examined. The AI system could potentially speed up the examination process, which could improve the effectiveness of cancer treatments and allow uropathologists to focus on other cases.
AI is becoming more common in the medical industry
More validation is needed before the AI system can be used in clinical practice. Researchers are currently planning to span the study across nine European countries with an expected completion date of December 2020. The goal of the study is to train the AI system to recognize cancer in biopsies taken from different laboratories with different digital scanners and with different growth patterns.
But Eklund’s AI system isn’t the only AI that’s being developed to help with cancer diagnoses. Google Health researchers published a paper in late 2019 evaluating the use of AI for breast cancer prediction in mammogram images.
Just like in Eklund’s study, the AI system performed just as well, if not better, than radiologists. But also just like in Eklund’s study, more validation is needed before the AI can be used in clinical practice.
Other types of AI that are currently being developed or are already being used by the medical industry include:
- PathAI: A developing machine learning technology that will assist pathologists in making more accurate cancer diagnoses. The technology has been used by drug developers like Bristol-Myers Squibb.
- Enlitic: A developing deep learning medical tool designed to streamline radiology diagnoses. The system analyzes unstructured medical data like patient medical history, blood tests, genomics, and radiology images to give doctors better insight into a patient’s needs in real-time.
- AI-enhanced microscopes: Doctors at Harvard University’s teaching hospital, Beth Israel Deaconess Medical Center, are using AI-enhanced microscopes to scan for harmful bacterias in blood samples. The CDC currently recommends removing old bandages and checking for signs of infection every 24 hours, but AI-enhanced microscopes could help to identify and predict harmful bacteria in blood with 95% accuracy. The AI could also be used in the future to catch other medical issues like prediabetes, which affects 33.9% of adults in the United States.
AI technology isn’t being developed to replace doctors but to support medical professionals by providing faster service and more accurate diagnoses. Saving time can mean saving lives in the medical industry, which makes machine learning not only transformative for medicine but also for patients.
