Large language models are quickly becoming a key building block in new information systems for administrative staff and clinicians in hospitals and healthcare systems. This form of artificial intelligence can perform tasks that no human can imagine.
Harrison.ai develops artificial intelligence technology to accelerate clinical diagnosis and offers a range of AI radiology and pathology tools. They are designed to improve physician efficiency to tackle burnout.
Healthcare IT News spoke with Dr. Aengus Tran, co-founder and CEO of Harrison.ai, about LLMs and radiology: why they’re a good match, what genAI models can do for radiologists, how radiologists can be assured of their quality and accuracy—and how adopting LLMs in radiology can help address the radiologist shortage.
Q. Why is a large language model suitable for radiology?
A. Large language models have the potential to address some of radiology’s most pressing challenges. While many of the AI models that have entered healthcare are only capable of performing predefined tasks, advances in machine learning are improving the ability of new models to learn continuously and generalize to domains in which the model has not been trained.
This is a transformative next step for AI in healthcare – a sector where drawing conclusions based on past experience and knowledge in the face of new and unfamiliar circumstances is critical to providing patients with the right care.
The way radiology LLMs are trained is no different than how medical students learn diagnostic radiology: through constant practice, reviewing cases, and studying the literature. A well-trained LLM model should be able to achieve human-like performance on tasks such as parsing radiology images to detect abnormalities, localizations, comparing priors, and predicting outcomes.
LLMs can provide an immediate and direct benefit for radiologists, as it helps them address the rapid growth of medical data by quickly processing and integrating information from multiple sources.
Whether interpreting textual data such as medical literature and patient histories or analyzing visual image data, these models can provide radiologists with rich insights that previously took significant time and resources to compile.
In addition, radiological images have been digitized, providing a wealth of high-quality, standardized data that is unique to the field and lends itself well to AI intervention.
Q. What can an LLM do for a radiologist?
A. Medical institutions around the world are struggling with an increasing amount of medical images and associated case-by-case data, a shortage of radiologists, and the risk of physician burnout.
An LLM specifically focused on radiology can quickly process medical information, patient histories, and imaging data, providing radiologists with comprehensive insights in a fraction of the time.
In addition, LLMs can help provide diagnostic decision support for radiologists by interpreting image data, identifying abnormalities, suggesting possible diagnoses, and automating time-consuming administrative tasks. Radiologists can then make faster and more accurate decisions, allowing them to see more patients while reducing their overall workload.
Contrary to previous concerns about AI replacing radiology jobs, LLMs – or at least the way we see them developing – not intended to replace human expertise, but to enhance and extend it.
Although many LLMs are globally influential, they have a broad, generic focus.
These generalist models are not suitable for a field that is completely dependent on accuracy and cannot accept errors. A specialized and highly nuanced function such as health care requires a specialized model.
Q. How can a radiologist be assured of the quality and accuracy of the work an LLM does for them? How can they feel comfortable?
A. A model is only as good as the data it is trained on – and we need to be sensitive to the risks and challenges that come with using LLMs. The effectiveness of LLMs depends on three key elements of their training data: quality, volume, and diversity. By leveraging datasets that excel in these aspects, we can create advanced systems capable of generating accurate, high-quality output.
Furthermore, a comprehensive evaluation is essential. Evaluating LLMs for use in radiology poses additional challenges: to evaluate fundamental models, we need to move to a paradigm in which we test them for their ability to recognize individual pathologies and their radiology interpretation skills in general.
This means that even more rigorous testing for safety and accuracy for LLMs is needed. This includes testing against international standards and benchmarks, where the performance of other LLMs in the industry, and subjecting the models to real-world assessments.
Several benchmarks have been introduced to evaluate and compare the performance of multimodal fundamental models on medical tasks. We believe that LLMs should not only be assessed against these benchmarks, but also against examinations taken by radiologists, which are considered the gold standard when it comes to interpreting medical images.
This rigorous evaluation process serves a dual purpose: it builds confidence in radiologists by demonstrating thorough validation of the model and simultaneously confirms its legitimacy as a reliable supporting technology.
Q. How can the introduction of LLMs in radiology contribute to solving the shortage of radiologists?
A. Global healthcare faces multiple intersecting challenges, including increasing volumes of imaging and associated case data, shortages of clinical professionals, and the risk of burnout for the remaining workforce. LLMs may help address these issues by improving productivity and efficiency in diagnostic processes:
-
They can improve the efficiency of manual data annotation to create large, labeled datasets for comprehensive medical imaging with artificial intelligence (AI).
-
They can provide easy access and retrieval of cases by analyzing radiology reports, enabling fast, efficient and continuous quality assessment.
-
Importantly, as a model that can work anywhere and at any time of day, LLMs can facilitate greater access to radiology services in underserved and remote areas. This can mean providing preliminary readings and support to clinicians who may be working in isolated locations or in resource-limited facilities, improving equitable access to timely and accurate diagnoses for patients around the world.
These are often time-consuming activities that can be streamlined by AI, allowing radiologists to focus on the critical decision-making elements of their work that have the greatest impact on patient care.
Follow Bill’s HIT reporting on LinkedIn: Bill Siwicki
Send him an email: bsiwicki@himss.org
Healthcare IT News is a publication of HIMSS Media.