A new analysis shows why models fall short in practice, how liability and equity issues slow adoption, and what lab leaders should consider as AI becomes a growing part of diagnostic workflows.
Artificial intelligence (AI) has made notable advances in medical imaging, but radiologists are not being displaced. For laboratory and diagnostic leaders, a recent analysis in Works in Progress highlights why AI has not replaced human expertise in radiology—and what this means for managing technology adoption in labs and hospitals.
In 2016, AI pioneer Geoffrey Hinton declared that “people should stop training radiologists now.” Since then, more than 700 FDA-cleared radiology AI models have entered the market, covering everything from stroke detection to lung cancer screening.
Companies such as Annalise.ai, Lunit, Aidoc, and Qure.ai offer tools that can identify dozens of diseases across modalities, reorder worklists, or generate structured draft reports. “On paper, radiology looks like the perfect target for automation,” the article noted, citing its reliance on digital images, pattern recognition, and quantitative benchmarks. Yet demand for radiologists has never been higher. In 2025, US residency programs offered a record 1,208 positions, and vacancy rates remain high as well.
Why Hasn’t AI Taken Over?
For leaders overseeing diagnostic services, three key elements are why AI has not replaced radiologists.
First, models struggle in real-world deployment. “Performance can drop by as much as 20 percentage points” when systems trained on narrow datasets are applied across different scanners, imaging protocols, or patient populations, the article explained. What works in a benchmark test may falter in a hospital with diverse workflows.
Second, liability and regulatory hurdles remain high. Assistive models that require physician review face fewer barriers, but autonomous systems must self-abort on poor image quality, identify unfamiliar equipment, and withstand rigorous scrutiny. Insurers have also drawn hard lines: one malpractice policy states that “coverage applies solely to interpretations reviewed and authenticated by a licensed physician; no indemnity is afforded for diagnoses generated autonomously by software.” Another bluntly imposes an “Absolute AI Exclusion.” For labs, this underscores the importance of risk management before deploying AI tools.
Photo credit: “Artificial Intelligence – Resembling Human Brain” by deepakiqlect is licensed under CC BY 2.0. To view a copy of this license, visit https://creativecommons.org/licenses/by/2.0/?ref=openverse.
Photo credit: “Cancer” by davis.steve32 is licensed under CC BY 2.0. To view a copy of this license, visit https://creativecommons.org/licenses/by/2.0/?ref=openverse.
Third, radiologists do much more than read scans. “Human radiologists spend a minority of their time on diagnostics and the majority on other activities, like talking to patients and fellow clinicians,” the commentary pointed out. Oversight of imaging protocols, interdisciplinary consultations, and patient communication all fall outside the reach of algorithms. Even as AI improves, demand for imaging may increase rather than decrease—a version of the Jevons paradox where greater efficiency leads to higher use. “The better the machines, the busier radiologists have become,” the article observed.
For laboratory leaders, the takeaway is not to fear replacement but to prepare for integration. AI tools are proving valuable in triaging urgent cases, flagging abnormalities, and drafting reports, but they remain narrow in scope—stroke, lung cancer, and breast lesions account for about 60% of models, yet represent only a fraction of total imaging work. As the article concluded, “Models can lift productivity, but their implementation depends on behavior, institutions and incentives.”
The challenge for labs is to create environments where AI augments human expertise rather than attempts to replace it. That means aligning technology adoption with clinical needs, providing training for staff, and working with insurers and regulators to ensure coverage and compliance.
For now, radiologists and the labs that support them are not going away. They are adapting, and AI will be a partner in that evolution.
New program draws bipartisan criticism and concern from patients and doctors.
Shrewd labs will keep an eye on the latest Centers for Medicare & Medicaid Services (CMS) prior authorization pilot that leans on artificial intelligence (AI) to determine treatment options for Medicare patients. While the Wasteful and Inappropriate Service Reduction Model pilot (WISeR) doesn’t directly mention lab tests, staying on the pulse of this growing trend will keep labs thinking ahead on how to minimize impact on bottom line, paperwork, and workflows when these pilots infiltrate lab testing.
An article from POLITICO reported that CMS will start a pilot version of the program as early as January 2026 in six states including Ohio, Texas, Oklahoma, Ariz., N.J., and Wash. Private AI companies will assist and focus on “services that have been vulnerable to fraud, waste and abuse in the past,” the article noted. The voluntary model is slated to span six years through December 31, 2031, according to the Centers for Disease Control and Prevention (CDC).
Among the types of procedures encumbered by the pilot program are knee arthroscopy for osteoarthritis, skin and tissue substitutions, and electrical nerve stimulator implants, CMS noted. All outpatient and emergency services would currently be excluded, they added, as well as “services that would pose a substantial risk to patients if substantially delayed.”
“All recommendations for non-payment will be determined by appropriately licensed clinicians who will apply standardized, transparent, and evidence-based procedures to their review,” CMS added.
The premise of the pilot is to eliminate wasteful spending, with CMS citing 25% of US healthcare spending falling in this category. “According to the Medicare Payment Advisory Commission Medicare spent up to $5.8 billion in 2022 on unnecessary or inappropriate services with little to no clinical benefit,” their website noted.
A Sour Reception
The pilot program is receiving a less-than-warm welcome from both parties—doctors, and patients alike, Politico noted. “It’s been referred to as the AI death panel. You get more money if you’re that AI tech company if you deny more claims. That is going to lead to people getting hurt,” Greg Landsman (D-Ohio) said during the committee hearing.
Landsman noted in the article from POLITICO that a bipartisan desire to put a halt to the program exists among growing concerns about patient harm coming from the program. Landsman “called for the program to be shut down until an independent review board could be erected to review the liability questions and ensure the AI prior authorization pilot doesn’t harm patients.”
“I’m concerned that this AI model will result in denials of lifesaving care and incentivize companies to restrict care,” Frank Pallone (D-N.J.) and House Energy and Commerce Committee ranking member said at the subcommittee meeting on the use of AI in health care held on Sept. 3.
“We have pretty good evidence that prior authorization as a process itself is fraught, adding that AI’s ability to improve the process for patients remains unproven,” Michelle Mello, Stanford University health law professor and witness at the hearing, said.
Looking Ahead
The involvement of AI in healthcare will only continue, and learning what aspects positively impact healthcare versus cause damage will continue to evolve.
Worth noting, there are already two unrelated lawsuits, against UnitedHealthcare and Cigna, that challenge the safety of AI use to deny patient care, POLITICO noted in the article.
Laboratory leaders should keep their eyes open and their ears to the ground on not only the pilot but all AI healthcare trends.
Recent laws in California, Utah, and Texas define new compliance standards for clinical laboratories employing AI in diagnostic and clinical messaging.
When it comes to oversight of artificial intelligence (AI) use in clinical laboratory, it behooves lab leaders to watch what is happening on the state level. In some cases, disclosure of AI use is a threshold states are monitoring.
For example, California Assembly Bill 3030, which went into effect Jan. 1, 2025, mandates transparency when generative AI is used in healthcare. Any health facility, laboratory, clinic, physician’s office, or group practice that employs generative AI to create patient communications about clinical information must include:
A prominent disclaimer stating the content was AI-generated.
Clear instructions that inform patients how to speak directly with a human clinician.
If a licensed provider reviews and approves the AI-generated communication, these requirements are waived. AB 3030 applies only to clinical—not administrative—messages. Non‑compliance can result in disciplinary actions from state regulators.
Laboratories using AI in patient-facing contexts should ensure their workflows include AI‑disclaimers, human‑review triggers, and clear ways for patients to contact providers.
“Symposium Cisco Ecole Polytechnique 9-10 April 2018 Artificial Intelligence & Cybersecurity” by Ecole polytechnique / Paris / France is licensed under CC BY-SA 2.0.
AI Disclosure in Utah
Meanwhile, Utah Senate Bill 226 updates its Artificial Intelligence Policy Act, tightening rules around how healthcare entities—including clinical labs—use generative AI in patient interactions. The rules went into effect May 7, 2025.
Under the state’s law, labs must disclose AI use only when:
A patient explicitly asks whether they’re interacting with AI, or
The lab uses AI in high-risk communications, such as delivering test interpretations, diagnostic results, or clinical advice.
Routine AI use in back-end operations or non-clinical messaging does not require disclosure.
A safe harbor provision protects labs from penalties if the AI system clearly identifies itself as non-human at the beginning and throughout the interaction.
Labs that use AI-generated content in patient portals, chatbots, or outreach must ensure compliance or face consumer protection penalties.
New Texas Law on AI
Texas passed a law in June that goes into effect Sept. 1, 2025, the regulates how AI is used within electronic health records (EHRs).
According to the law, providers that use AI for recommendations on diagnosis or treatment based on a patient’s medical record must review all information obtained through AI to ensure its accuracy before entering the information into a patient’s EHR.
The law also “imposes a strict data localization mandate, prohibiting the physical offshoring of electronic medical records,” law firm Holland & Knight noted. “This requirement applies not only to records stored directly by healthcare providers but also to those maintained by third-party vendors or cloud service providers.”
As the cancer registry expands it will increasing become more useful to anatomic pathologists, histopathologists, oncologists, and even clinical laboratories
Oncologists, histopathologists, anatomic pathologists, and other cancer physicians now have a powerful new Wikipedia-style tumor registry to help them with their diagnoses and in educating patients on their specific types of cancer. Clinical laboratory managers may find it useful to understand the value this searchable database, and it can help their staff pathologists as well.
Free to use by both physicians and patients the World Tumor Registry (WTR) is designed “to minimize diagnostic errors by giving doctors a searchable online database of cancers that have been collected and categorized with cellular images collected from around the world,” Pittsburg-Post Gazette reported.
Prompt, accurate cancer diagnoses offer cancer patients the best chance for optimal treatment outcomes. However, many medical professionals around the globe do not have the training and resources to offer superior cancer diagnoses. That deficiency can translate to inferior treatment options and lower survival rates among cancer patients.
To help improve cancer diagnoses, pathologist Yuri E. Nikiforov, MD, PhD, Division Director, Molecular and Genomic Pathology, Vice Chair of the Department of Pathology, and Professor of Pathology, University of Pittsburgh, developed the WTR to provide educational and practical resources for individuals and organizations involved in cancer research.
Officially announced at the United States and Canadian Academy of Pathology (USCAP) annual convention, the WTR is an open-access catalog of digital microscopic images of human cancer types and subtypes.
The lower cost of technology and improved speed of access via the internet are technologies enabling this effort.
“We are creating sort of a Wikipedia for cancer images,” said Alyaksandr V. Nikitski, MD, PhD (above), Research Assistant Professor of Pathology, Division of Molecular and Genomic Pathology at Pittsburg School of Medicine and Administrative Director of the WTR, in an exclusive interview with Dark Daily. “Anyone in the world, if they can access the internet, can look at the well-annotated, diagnostic digital slides of cancer,” said Nikitski. Clinical laboratories may also find this new pathology tool useful. (Photo copyright: Alyaksandr V. Nikitski)
Minimizing Diagnostic Errors
Based in Pittsburgh, the WTR is freely available to anyone for viewing digital pathology slides of known cancer tumors as well as borderline and questionable cases. On the website, individuals can search for pictures of tumors in the registry by diagnosis, specific cohorts, and by microscopic features. Individuals may search further by tumor type and subtype to receive a picture of related tumors.
According to the WTR website, the mission of the nonprofit “is to minimize diagnostic errors, eliminate inequality in cancer recognition, diagnosis, and treatment in diverse populations, and improve outcomes by increasing access to the diagnostic pathology expertise and knowledge of microscopic characteristics of cancers that occur in different geographic, environmental, and socio-economic settings.”
This new comprehensive initiative will eventually encompass cancer images from all over the world.
“Let’s assume that I am a pathologist or a trainee who has little experience, or I don’t have access to collections of atypical tumors,” Nikitski explained. “I can view tumor collections online [in the WTR database] and check how typical and rare tumors look in various geographic regions and environmental settings.”
Once an image of a slide is selected, users will then receive a brief case history of the tumor in addition to such data as the age of the patient, their geographic location, sex, family history of the disease, and the size and stage of the tumor.
Increasing Probability of Correct Diagnosis
Pathologists and clinicians may also predict the probability of a particular diagnosis by searching under the microscopic feature of the database. This feature utilizes an innovative classifier known as PathDxFinder, where users may compare a slide from their lab to slides in the database by certain criteria. This includes:
After completing the questions above, the user presses the “predict diagnosis” button to receive the probability of cancer and most likely diagnosis based on the answers provided in the questionnaire.
WTR Editorial Boards
The WTR represents collections for each type of cancer site, such as lung or breast. A chairperson and editorial board are responsible for reviewing submitted slides before they are placed online. The editorial boards include 20 pathologists who are experts in diagnosing cancer categories, Nikitski explained.
Thousands of identified microscopic whole slide images (WSI) representing various types of cancer are deposited by the editors and other contributors to the project. The editorial board then carefully analyzes and compiles the data before posting the images for public viewing.
The editorial boards are located in five world regions:
Africa and the Middle East
Asia and Oceania
Central and South America
North America and Europe
Northern Asia
Any physicians or pathologists can contribute images to the database, by “simply selecting the editor of their region on the website, writing their name, and asking if they can submit tumor cases,” Nikitski stated.
“We have established a platform that allows pathologists to contact editors who are in the same geographic region,” he added.
Helping Physicians Identify Cancer Types
In a YouTube video, Nikiforov states that the WTR is an “educational nonprofit organization rooted in [the] beliefs that every cancer patient deserves accurate and timely diagnosis as the first and essential step in better treatment and outcomes.”
“We believe this can be achieved only when modern diagnostic tools and technologies are freely available to every physician and pathologist. Only when we understand how microscopic features of cancer vary in different geographic, environmental and ethnic populations, and only by integrating histopathology with clinical immunohistochemical and molecular genetic information for every cancer type,” he stated.
Since patient privacy is important, the database contains only basic data about patients, and all patient information is protected.
Launched in March, there are currently more than 400 thyroid tumor slides available to view in the online database. At the time of the announcement, the WTR platform was planned to be implemented in three phases:
Thyroid cancer (released in March of this year).
Lung cancer and breast cancer (anticipated to be completed by the third quarter of 2026).
Remaining cancers, including brain, soft tissue and bone, colorectal, head and neck, hematolymphoid, female genital, liver, pancreatic, prostate and male genital, skin, urinary system, pediatric, other endocrine cancers, and rare cancers (anticipated to be completed by the end of 2029).
“We believe that this resource will help physicians and pathologists practicing in small or big or remote medical centers to learn how cancer looks under a microscope in their own communities,” Nikiforov said in the video. “We also see WTR as a platform that connects physicians and scientists from different parts of the world who can work together to better understand and treat cancer.”
Catalogs like the World Tumor Registry might potentially create a pool of information that that could be mined by analytical and artificial intelligence (AI) platforms to ferret out new ways to improve the diagnosis of certain types of cancer and even enable earlier diagnoses.
“It is an extremely useful resource,” Nikitski said.
Anatomic pathologists will certainly find it so. And clinical laboratory managers may find the information useful as well when interacting with histopathologists and oncologists.
The deal will enable Crosscope’s digital pathology platform to layer around Clarapath’s histology automation hardware, a combination that could improve quality and efficiencies in diagnostic services for future customers, according to a Clarapath press release.
Clarapath’s goal with its products is to automate certain manual processes in histology laboratories, while at the same time reducing variability in how specimens are processed and produced into glass slides. In an exclusive interview with Dark Daily, Eric Feinstein, CEO and President at Clarapath said he believes the resulting data about these activities can drive further changes.
“A histotechnologist turns a microtome wheel and makes decisions about a piece of tissue in real time,” noted Feinstein, who will speak at the Executive War College on Diagnostics, Clinical Laboratory, and Pathology Management on April 25-26 in New Orleans. “All of that real-time data isn’t captured. Imagine if we could take all of that data from thousands of histotechnologists who are cutting every day and aggregate it. Then you could start drawing definitive conclusions about best practices.”
“Clarapath’s foundation is about creating consistency and standardizing steps in histology—and uncovering the data that you need in order to accomplish those goals as a whole system,” Eric Feinstein (above), CEO and President at Clarapath told Dark Daily. “A histology lab’s workflow—from when the tissue comes in to when the glass slide is produced—should all be connected.” Many processes in histology and anatomic pathology continue to be manual. Automated solutions can contribute to improved productivity and reducing variability in how individual specimens are processed. (Photo copyright: Clarapath.)
Details Behind Clarapath’s Deal to Acquire Crosscope
As part of its acquisition, Clarapath of Hawthorne, New York, has retained all of Crosscope’s employees, who are located in Mountain View, California, and Bombay, India. Financial terms of the deal were not disclosed.
Clarapath’s flagship histology automation product is SectionStar, a tissue sectioning and transfer system designed to automate inefficient and manual activities in slide processing. The device offers faster and more efficient sample processing while reducing human involvement. Clarapath expects SectionStar be on the market in 2023. The company is currently taking pre-orders.
Meanwhile, Crosscope developed Crosscope Dx, a turnkey digital pathology solution that provides workflow tools and slide management as well as AI and machine learning to assist pathologists with their medical decision-making and diagnoses.
Adoption of Digital Pathology and Automation Can Be Challenging
Digital pathology has experienced growing popularity in the post-COVID-19 pandemic world. This is not only because remote pathology case reviews have become increasingly acceptable to physicians but also because of the ongoing shortages in clinical laboratory staffing.
“A pain point today for clinicians and laboratories is labor. That’s across the board,” Feinstein said. “We can help solve that with SectionStar.”
Feinstein does not believe adoption of digital pathology and histology automation is proceeding slowly, but he does acknowledge barriers to healthcare organizations installing the technologies.
“There are lots of little things that—from a workflow perspective—people have outsized expectations about,” he explained. “Clinicians and administrators are not used to innovating in a product sense. They may be innovating on how they deliver care or treatment pathways, but they’re not used to developing an engineering product and going through alpha and beta stages. That makes adopting new technology challenging.”
Medical laboratory managers and pathologists interested in pursuing histology automation and digital pathology should first determine what processes are sub-optimal or would benefit from the standardization hardware and software can offer. Being able to articulate those gains can help build the case for a return on investment to decision-makers.
Another resource to consider: Feinstein will speak about innovations for remote histology laboratory workers at the upcoming Executive War College for Clinical Laboratory, Diagnostics, and Pathology Management on April 25-26 in New Orleans. His session is titled, “Re-engineering the Classic Histology Laboratory: Enabling the Remote Histotechnologist with New Tools That Improve Productivity, Automate Processes, and Protect Quality.”
Though smartphone apps are technically not clinical laboratory tools, anatomic pathologists and medical laboratory scientists (MLSs) may be interested to learn how health information technology (HIT), machine learning, and smartphone apps are being used to assess different aspects of individuals’ health, independent of trained healthcare professionals.
The issue that the Cedars Sinai researchers were investigating is the accuracy of patient self-reporting. Because poop can be more complicated than meets the eye, when asked to describe their bowel movements patients often find it difficult to be specific. Thus, use of a smartphone app that enables patients to accurately assess their stools in cases where watching the function of their digestive tract is relevant to their diagnoses and treatment would be a boon to precision medicine treatments of gastroenterology diseases.
“This app takes out the guesswork by using AI—not patient input—to process the images (of bowel movements) taken by the smartphone,” said gastroenterologist Mark Pimentel, MD (above), Executive Director of Cedars-Sinai’s Medically Associated Science and Technology (MAST) program and principal investigator of the study, in a news release. “The mobile app produced more accurate and complete descriptions of constipation, diarrhea, and normal stools than a patient could, and was comparable to specimen evaluations by well-trained gastroenterologists in the study.” (Photo copyright: Cedars-Sinai.)
Pros and Cons of Bristol Stool Scale
In their paper, the scientists discussed the Bristol Stool Scale (BSS), a traditional diagnostic tool for identifying stool forms into seven categories. The seven types of stool are:
Type 1: Separate hard lumps, like nuts (difficult to pass).
Type 2: Sausage-shaped, but lumpy.
Type 3: Like a sausage, but with cracks on its surface.
Type 4: Like a sausage or snake, smooth and soft (average stool).
Type 5: Soft blobs with clear cut edges.
Type 6: Fluffy pieces with ragged edges, a mushy stool (diarrhea).
Type 7: Watery, no solid pieces, entirely liquid (diarrhea).
But even with the BSS, things can get murky for patients. Inaccurate self-reporting of stool forms by people with IBS and diarrhea can make proper diagnoses difficult.
“The problem is that whenever you have a patient reporting an outcome measure, it becomes subjective rather than objective. This can impact the placebo effect,” gastroenterologist Mark Pimentel, MD, Executive Director of Cedars-Sinai’s Medically Associated Science and Technology (MAST) program and principal investigator of the study, told Healio.
Thus, according to the researchers, AI algorithms can help with diagnosis by systematically doing the assessments for the patients, News Medical reported.
30,000 Stool Images Train New App
To conduct their study, the Cedars-Sinai researchers tested an AI smartphone app developed by Dieta Health. According to Health IT Analytics, employing AI trained on 30,000 annotated stool images, the app characterizes digital images of bowel movements using five parameters:
BSS,
Consistency,
Edge fuzziness,
Fragmentation, and
Volume.
“The app used AI to train the software to detect the consistency of the stool in the toilet based on the five parameters of stool form, We then compared that with doctors who know what they are looking at,” Pimentel told Healio.
AI Assessments Comparable to Doctors, Better than Patients
According to Health IT Analytics, the researchers found that:
AI assessed the stool comparable to gastroenterologists’ assessments on BSS, consistency, fragmentation, and edge fuzziness scores.
AI and gastroenterologists had moderate-to-good agreement on volume.
AI outperformed study participant self-reports based on the BSS with 95% accuracy, compared to patients’ 89% accuracy.
Additionally, the AI outperformed humans in specificity and sensitivity as well:
Specificity (ability to correctly report a negative result) was 27% higher.
Sensitivity (ability to correctly report a positive result) was 23% higher.
“A novel smartphone application can determine BSS and other visual stool characteristics with high accuracy compared with the two expert gastroenterologists. Moreover, trained AI was superior to subject self-reporting of BSS. AI assessments could provide more objective outcome measures for stool characterization in gastroenterology,” the Cedars-Sinai researchers wrote in their paper.
“In addition to improving a physician’s ability to assess their patients’ digestive health, this app could be advantageous for clinical trials by reducing the variability of stool outcome measures,” said gastroenterologist Ali Rezaie, MD, study co-author and Medical Director of Cedars-Sinai’s GI Motility Program in the news release.
The researchers plan to seek FDA review of the mobile app.
Opportunity for Clinical Laboratories
Anatomic pathologists and clinical laboratory leaders may want to reach out to referring gastroenterologists to find out how they can help to better serve gastro patients. As the Cedars-Sinai study suggests, AI smartphone apps can perform BSS assessments as good as or better than humans and may be useful tools in the pursuit of precision medicine treatments for patient suffering from painful gastrointestinal disorders.
