Some artificial intelligence experts have openly worried that recent high-end AI programs are starting to become too good at surpassing humans – which could erode trust in the technology.
Peter Shen, head of digital business and automation at Siemens Healthineers, worries about the risk of distrust that could pose to AI in healthcare in particular – an industry that has been slow to adopt the technology. this despite the potential to improve patient care. Among other things.
We interviewed Shen to dig deeper into the world of healthcare AI and discuss what needs to be done to ensure trust and wider adoption of the technology.
Q. How could a program like ChatGPT change healthcare in the future?
ONE. Ever since artificial intelligence company OpenAI made ChatGPT – a free, interactive chatbot powered by machine learning – available to the public, this interactive tool has made headlines. It can answer questions, compose emails, and create lengthy written content (including student stories, articles, and essays) with incredible speed, as well as suggestive levels of quality. for a human author.
The implications of this technology, which is seen as a paradigm-shifter in the evolution of AI, is wide-ranging. These include potential applications in marketing and journalism as well as healthcare. Concerns have also arisen regarding the possible abuse of ChatGPT by students, as well as its credibility with some of the sources it cites.
Some critics of the potential use of ChatGPT in healthcare argue that clinicians would use such a tool to provide a critical patient diagnosis or clinical solution. However, I see a solution like this that can offer the benefit of potentially consuming large amounts of textual data and summarizing the results.
That capability is particularly relevant in the healthcare sector, where clinicians are faced with the challenge of absorbing huge amounts of emerging digital data in the form of publications and research. science related to their expertise. Armed with information about new, potentially relevant clinical outcomes, clinicians can change the diagnosis and treatment paradigm for patients.
ChatGPT represents an opportunity to consume and summarize this scientifically publicized data, keeping the physician up to date. For maximum effectiveness, ChatGPT will need to be provided with the latest publications on an ongoing basis. This practical application of the tool will benefit clinicians and the healthcare industry at large while avoiding AI making definitive clinical decisions, as trust on that front remains. still difficult to grasp.
Q. Healthcare is an industry slow to adopt AI despite the potential to improve patient care. Why do you think this is?
ONE. In its early days, AI was the subject of considerable speculation and its actual value has yet to be established. One thing seems certain, if you believe the original headlines and professional polls: It will displace radiologists and other clinicians. Perceiving a new technology as a job threat is always counterproductive to adoption.
However, in recent years, concerns about that alternative have subsided, and the healthcare community has gained a better understanding of the core benefits of AI. In radiology, this technology has been shown to have the potential to not only save time by automating repetitive tasks, but also identify overlooked areas of interest by taking advantage of sample form.
In addition, AI has begun to provide qualitative and instructive visualizations related to suspected malignancies.
But new questions have arisen. If AI introduces additional data points that may not even be relevant to the diagnosis, how much time does it actually save the clinician? And how can this new data be contextualized and incorporated into a single report type? Also, how do we solve the implicit bias problem, where algorithms are trained on batches of data that do not include differences in gender, race, and geography?
These questions highlight how much AI still has to evolve to provide broader, more substantive value to not only clinicians but also patients and healthcare facilities. Until those questions are answered, some healthcare organizations will struggle with justifying the costs and conflicts with broader AI adoption.
Q. How can the healthcare industry increase AI adoption?
ONE. For adoption to grow, AI must make a difference in the eyes of radiologists and other clinicians when it can offer something truly new. After all, these professionals already know how to make diagnostic or therapeutic decisions; They’ve done it all their careers.
They need to be able to determine if additional AI information is useful, relevant, and worth considering. They need to know how it changes their diagnosis – if at all.
Developing AI models that include the rationale for their findings will make this tool more useful to radiologists and other clinicians, helping them become champions have more say in AI. Those clinicians will also feel more confident about using AI now that its implicit bias has been removed through next-generation algorithms fed with continuous data representing groups. diverse patients.
Even more important to AI adoption is the seamless, efficient integration of AI’s complementary data into routine clinical workflows. The additional information provided by AI should simply be a more accessible tool that complements the clinician’s established routine; it must never be intrusive with respect to that habit.
But perhaps the bigger, more overarching challenge for AI involves changing our collective thinking about what AI’s role in healthcare is and isn’t. The common perception in some circles is that implementing AI will lead to the tool, rather than the clinician, making decisions.
Even Siemens Healthineers stand-alone AI solutions are companion technologies designed to assist the clinician who makes the final decisions regarding patient care. Fully realizing that AI does not – and should not – bear the burden of making clinical decisions is key to wider acceptance.
Q. Do you have a vision for the next generation of health AI? Changes we can see in the future?
ONE. Currently, we use AI to detect potential abnormalities on chest CT images. Taking AI to the next level involves using a multi-data middleware platform to combine that kind of visual information with other forms of healthcare data ever before – in-room diagnostics experiments, pathology results, genetic information – and overlay AI over those silos to find correlations.
This use of AI will lead to more informed diagnoses and more personalized treatment decisions.
A hypothetical example: A urologist, based on professional experience, might prescribe radiation therapy for 10 weeks, three times a week, to treat a patient with prostate cancer. But if that urologist can examine all forms of data available from that patient — imaging, lab, pathology, and genomics — and overlay AI to find correlations in that data, the result can be a personalized treatment plan proposed with a miniaturized regimen.
It may include only five weeks of irradiation just once or twice a week.
This type of AI-assisted personalized treatment planning has huge implications for patient care. Organizations can apply it to entire groups of patients with similar characteristics to greater success. Using artificial intelligence in this way demonstrates real population health management, which is the goal of Siemens Healthineers.
Q. How can AI provide broader value to the healthcare system as a whole?
ONE. If AI can not only assist in making accurate diagnosis and treatment decisions for each patient, but also scale that personalized medicine approach to affect entire patient groups, it will prove value to the health care system beyond a particular specialty or specialty.
A key component of that approach is an integrated data management layer that can bring together different types of information and bring them into one platform to inform planning and prescribing. Now, a number of progressive healthcare organizations are moving in that direction.
In a related direction, AI may one day prove its broader value by creating highly accurate models of a patient’s anatomical structure. These models can demonstrate, non-invasively, how that anatomy responds to different forms of treatment.
Ultimately, that “digital copy” of the patient will non-invasively assist the clinician in determining the optimal personalized treatment. More broadly, it will also allow organizations to place patients in a health-focused environment and help identify ways to keep that person healthy. AI’s ability to deliver that benefit could completely transform healthcare.
Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email the writer: [email protected]
Healthcare IT News is a publication of HIMSS Media.
