Editor: Deborah Sesok-Pizzini, MD, MBA, adjunct professor, Department of Clinical Pathology and Laboratory Medicine, Perelman School of Medicine, University of Pennsylvania, Philadelphia.
Ethical and regulatory perspectives on generative AI in pathology
March 2025—Generative artificial intelligence is now readily available and has created a plethora of interest in health care, including in pathology and laboratory medicine. Unlike traditional AI, generative AI (GenAI) doesn’t rely solely on historical data to make predictions but instead uses patterns within historical data to create entirely new data. Excitement over GenAI must be weighed against the reality of maintaining human control over the technology and interpreting the data. The use of GenAI in health care may include not only streamlining administrative tasks but also performing safety critical clinical functions, such as image-based diagnosis. The risk-based approach and quality management of GenAI must be managed individually for each situation. This creates ethical, legal, and social issues that should be considered in the scope of patient care. The promise of GenAI to create novel human thinking also creates challenges for quality management. The technology will reflect the same biases within the data sets that were used to train it. Therefore, accountability is necessary for each stage of the AI life cycle. Because GenAI models lack transparency, strategies are needed to educate patients and health care providers about the benefits and limitations of the technology. This will help build trust in the technology and encourage early adoption. The authors conducted a study to summarize the current frameworks for the ethical development and management of GenAI in health care. They analyzed scientific journals, organizational websites, and recent guidelines containing information about AI ethics and regulations. The authors first identified the major stakeholders and what they expect and require of GenAI in pathology and laboratory medicine. Ultimately, patients are the most important stakeholders and, in general, desire the newest technologies in medicine, but they have concerns about privacy violations and misdiagnosis. Patients have a right to know how their personal health data will be used and should be made aware of how to opt in or out of providing their data for AI training algorithms. They want their health data to be used fairly and transparently in GenAI. As patient advocates, health care providers may have some of the same concerns, as well as concerns about how GenAI can impact their profession and specific jobs. Pathologists may need to develop new skills, such as in the areas of data analysis and interpretation, to work effectively with AI systems. This may necessitate continuing education, mentorship programs, and cross-functional collaboration among health care professionals. The authors’ research showed that while numerous publications address the use of AI, it is still a new and evolving technology. The authors noted that effective and ethical use of GenAI requires processes and accountability within the technology industry, health care organizations, regulatory agencies, and professional societies. As AI continues to develop, several safety mechanisms for the responsible use of GenAI are needed to provide ethical oversight that maximizes benefits and mitigates risks. The authors concluded that GenAI applications need to be based on ethical principles that protect patient safety and patient rights. Efforts to develop such principles are still in the early stages. Health care organizations, medical associations, and regulatory agencies must collaborate to prioritize the development of an ethical and quality infrastructure to support the responsible use of GenAI.
Jackson BR, Rashidi HH, Lennerz JK, et al. Ethical and regulatory perspectives on generative artificial intelligence in pathology. Arch Pathol Lab Med. 2025;149(2):123–129.
Correspondence: Dr. Brian R. Jackson at brian.jackson@aruplab.com
Circulating nucleated red blood cells: an updated reference interval
Circulating nucleated red blood cells are not present in healthy people after the neonatal period. They are considered an abnormal finding and a sign of underlying disease in children and adults. Contemporary hematology analyzers can measure nucleated RBCs (nRBCs) at very low levels compared to manual methods using light microscopy, which are considered the gold standard for evaluating nRBCs. Any number of nRBCs detected by a Wright-stained peripheral smear is deemed abnormal. However, manual counting of nRBCs is labor intensive and prone to interobserver and intraobserver variability. Therefore, clinical laboratories are less likely to use manual methods to determine nRBCs. However, many laboratories still rely on reference intervals established using these manual methods. A large academic hospital clinical lab had initially validated its nRBC reference interval at 0.00 to 0.01 × 106/μL, using a Sysmex XE hematology analyzer. After upgrading to a Sysmex XN-10 analyzer, the lab verified and subsequently adopted this reference interval for reporting. Shortly after validation, the lab noticed an uptick in abnormally high nRBC results. The authors subsequently hypothesized that the Sysmex XN-10 hematology analyzer was more sensitive than the XE analyzer for detecting nRBCs in peripheral blood and that the previously established reference interval needed updating. Therefore, they conducted a study to determine if current reference intervals for nRBCs are clinically relevant. For the study, they performed a prospective analysis of 405,300 specimens from nonhospitalized individuals who received a complete blood count. The XN-10 was used to obtain the CBC results for all blood samples. Applying inclusion and exclusion criteria yielded a total of 66,498 specimens from those ages 18 to 69 years. The mean nRBC counts were compared between males and females using an unpaired t test (P<.001). Of the 66,498 samples with otherwise normal CBC results, 338 showed results that were abnormal and outside the established reference interval, and 336 of 66,498 had nRBC results greater than 0.01 × 106/μL. Two samples had nRBC values greater than 0.10 × 106/μL. NRBC values in excess of 0.10 × 106/μL were infrequent, regardless of patient age or sex. Linear regression analysis of nRBC counts based on age and sex showed that females have a 0.0002 × 106/μL lower nRBC count than males, and a t test showed a small but statistically significant difference (P<.001) of 0.0002 in the average nRBC count for male and female patients, where females have a lower average nRBC count than males. Reference intervals are usually derived from the typical distribution of results obtained from a healthy population and defined by results that include the central 95 percent of the reference population. However, reference intervals are only guidelines, and they should be used with other clinical and laboratory results to determine the next steps in patient care. Technological advances and more widespread use of automated hematology analyzers provide greater sensitivity and specificity for peripheral blood cell counts in a CBC. The increased sensitivity is due to analysis of much greater blood volume using the automated method. When the authors adjusted the reference interval from 0.01 × 106/μL to 0.10 × 106/μL, the number of abnormal CBC results dropped to two. This led to a decrease in patient referrals for elevated nRBCs with otherwise normal CBC results. The authors concluded that evaluating reference intervals with newer, more sensitive automated instruments is essential to limit abnormal flagging of results in otherwise healthy people. In this study, changing the nRBC reference intervals reduced referrals, patient anxiety, and costs.
Meredith AA, Meredith NR, Smith L, et al. Circulating nucleated red blood cells: an updated reference interval. Arch Pathol Lab Med. 2024;148:1365–1370.
Correspondence: Dr. Neil R. Meredith at nmeredith@wtamu.edu