Large Language Models in Diagnostic Pathology – Opportunities, Challenges, and Ethical Considerations
Artificial Intelligence (AI) technologies, driven mainly by deep learning (DL) algorithms, are becoming commonplace. Emergence and rapid evolution of large language models (LLMs), exemplified by ChatGPT from OpenAI, holds immense potential to augment current diagnostic pathology workflows. LLMs have not only sparked inquiries about their predictive potential, but also raised questions regarding the emergence of sentient AI. Integration of LLMs within digital pathology is an area of active research. This presentation will discuss the key challenges and barriers associated with the implementation and utilization of LLMs in the practice of pathology. A notable challenge associated with the use of AI models in medical practice lies in the lack of contextual understanding and interpretability associated with these algorithms. In the absence of a clear understanding of the predictive mechanisms of AI-LLMs, implementation is likely to be a challenge in regular medical/pathology practice. Speakers will shed light on topics such as the risks of biases in training data driving disparities and inaccuracies in diagnostic outcomes provided by the AI algorithms. The presentation will also address ethical aspects of LLMs, focusing on topics such as patient privacy, data security, and the responsible utilization of AI-LLMs in pathology. Integration of LLMs could potentially influence healthcare professionals' autonomy and decision-making process, requiring a deep and thoughtful discussion of the roles of AI-enabled automation and complementary human pathologist expertise. Issues of regulatory oversight will be examined, underlining the importance of robust guidelines and frameworks to ensure safe and ethical deployment of LLMs in pathology. The presentation aims to highlight the need for domain experts like pathologists to actively engage with, and steer the development of AI-LLMs in the practice of pathology. Collaborative efforts between AI developers and domain experts (like pathologists) is critical to navigate the manifold challenges and ethical complexities associated with AI-LLM adoption. AI technologies like LLMs are immensely powerful, and when harnessed appropriately, have the potential to dramatically change the way pathology is practiced in the future.
- Understand a rapidly developing area of AI technology, namely, large language models (LLMs) and its likely impact on pathology
- Review the critical role of domain experts, such as pathologists, in the development and implementation of ethical AI-LLM workflows
- Appreciate key issues of importance such as the role of data in driving biases of AI-LLM development and implementation.
Rama Gullapalli, MD, PhD
University of New Mexico Health Sciences Center
Rama Gullapalli, MD, PhD is a physician-scientist in the departments of Pathology, Chemical and Biological Engineering at the University of New Mexico (UNM). Dr. Gullapalli obtained his medical degree from the Armed Forces Medical College (AFMC) in India. He has a master’s degree in optical electrical engineering and a PhD in bioengineering from The Pennsylvania State University. Dr. Gullapalli completed his residency in Clinical Pathology and a fellowship in Molecular Genetic Pathology at the University of Pittsburgh Medical Center (UPMC). Dr. Gullapalli is interested in the convergence of technology with traditional pathology practice, with a focus on next generation sequencing (NGS), digital pathology, clinical informatics and personalized medicine. The Gullapalli research lab is focused on two research themes - 1) Hepatobiliary cancers with a focus on environment-microbiome-inflammation interactions 2) Understanding the role of environmental heavy metal pollutants (e.g., Cadmium) as a risk factor of hepatobiliary pathophysiology and incidence disparities.
Ehsan Ullah, MBBS, MPhil, PhD
Te Whatu Ora | Health New Zealand
Ehsan Ullah is a histopathologist turned biomedical scientist and laboratory manager.
Ehsan currently manages New Zealand’s largest Anatomical Pathology laboratories which provide diagnostic histology service to over 800 patients per day and additional over 500 cytology and another few hundred molecular (HPV) cancer screening tests per day. Ehsan also manages National Perinatal Pathology Service for New Zealand.
Ehsan explored the limitless opportunities of the application of artificial intelligence and automation in clinical diagnoses as part of his doctoral work. Ehsan is strong advocate of Digital Pathology and Computational Pathology and is contributing as member of the Northern Regional Digital Pathology working group for New Zealand, tasked with implementation of the technology in Northern Region and then nationally within public health system of New Zealand.