Artificial intelligence (AI) is changing medicine — including AI in ultrasound imaging, which appears to offer real promise for improving the efficiency of care and accuracy of diagnoses.
Some clinicians may still have qualms about using AI in medicine. But as AI becomes more commonplace in ultrasound imaging, Obstetricians and Gynecologists may wonder how they worked without it.
Does AI in Medicine Have Drawbacks?
One of the primary fears about AI across industries is that it will eliminate jobs. Forbes reported in 2019 that although this fear is common, AI systems are generally not designed to replace workers. Instead, they change the way employees work. AI is unable to "operate in a fully autonomous mode" — it can only function as a support system for a doctor or other professional who provides human oversight. These systems sort large amounts of data quickly, provide calculations and recognize patterns that humans cannot, but their main goal is to enhance physicians' capabilities.
In medicine, perhaps one of the most salient criticisms of AI is that it still relies on human programmers who are capable of committing errors: If the algorithms behind a diagnostic tool are biased, then the results are often biased. As Harvard Business Review wrote in 2019, AI systems are trained using data that contains information obtained by researchers, who may have made assumptions based on gender or other stereotypes. AI has the potential to uncover human bias, but this is still a potential problem worth acknowledging.
Applying AI to OB and GYN
Ultrasound is one of the most efficient and cost-effective modes of imaging, but its clinical utility is dependent on the capabilities of the clinician. Research published in Ultrasound in Obstetrics and Gynecology (UOG) reports that AI has not yet greatly impacted the field of OB/GYN, citing a need for collaborations between ultrasound users and AI developers in order to bring its vast capabilities into reality.
According to the UOG review, emerging AI has the potential to recognize what an ultrasound machine is "seeing" and advise the sonographer. Some of these capabilities include:
- Recognizing the boundaries of an anatomical structure.
- Instructing the sonographer how to move the probe to obtain a better image.
- Offering a diagnosis based on abnormalities in the image.
- Detecting standard planes.
- Measuring endometrial thickness and classifying ovarian cysts.
Researched published in the journal Cureus gives several more instances of how AI may soon assist obstetricians and gynecologists with decision making, case management and medical errors, noting that some of these are in development or limited use. Examples include fetal heart rate monitoring and screening for gestational diabetes. The National Cancer Institute describes an AI method of detecting cervical cancer, which was found to be superior to both visual inspection after vinegar rinse, and pap smear.
As advanced as AI is becoming, it will never be able to do everything a living, breathing physician can. It cannot explain a diagnosis to a patient with compassion, or decide on an individualized plan of care that accounts for a patient's wishes. It can, however, give clinicians additional information with more consistency, allowing for a more confident diagnosis. Quicker, more precise scans that offer better quality and reduce variability are one way of carving out more time for other tasks. The myriad ways AI in ultrasound can improve patient care have yet to fully emerge, but these technological breakthroughs are already showing benefits.
