Radiological method using AI algorithms to detect, locate, and identify cancer could negate the need for invasive, painful clinical laboratory testing of tissue biopsies
This will be of interest to histopathologists and radiologist technologists who are working to develop AI deep learning algorithms to read computed tomography scans (CT scans) to speed diagnosis and treatment of cancer patients.
“Researchers used the CT scans of 170 patients treated at The Royal Marsden with the two most common forms of retroperitoneal sarcoma (RPS)—leiomyosarcoma and liposarcoma—to create an AI algorithm, which was then tested on nearly 90 patients from centers across Europe and the US,” the news release notes.
The researchers then “used a technique called radiomics to analyze the CT scan data, which can extract information about the patient’s disease from medical images, including data which can’t be distinguished by the human eye,” the new release states.
The research team sought to make improvements with this type of cancer because these tumors have “a poor prognosis, upfront characterization of the tumor is difficult, and under-grading is common,” they wrote. The fact that AI reading of CT scans is a non-invasive procedure is major benefit, they added.
“This is the largest and most robust study to date that has successfully developed and tested an AI model aimed at improving the diagnosis and grading of retroperitoneal sarcoma using data from CT scans,” said the study’s lead oncology radiologist Christina Messiou, MD, (above), Consultant Radiologist at The Royal Marsden NHS Foundation Trust and Professor in Imaging for Personalized Oncology at The Institute of Cancer Research, London, in a news release. Invasive medical laboratory testing of cancer biopsies may eventually become a thing of the past if this research becomes clinically available for oncology diagnosis. (Photo copyright: The Royal Marsden.)
Study Details
RPS is a relatively difficult cancer to spot, let alone diagnose. It is a rare form of soft-tissue cancer “with approximately 8,600 new cases diagnosed annually in the United States—less than 1% of all newly diagnosed malignancies,” according to Brigham and Women’s Hospital.
In their published study, the UK researchers noted that, “Although more than 50 soft tissue sarcoma radiomics studies have been completed, few include retroperitoneal sarcomas, and the majority use single-center datasets without independent validation. The limited interpretation of the quantitative radiological phenotype in retroperitoneal sarcomas and its association with tumor biology is a missed opportunity.”
According to the ICR news release, “The [AI] model accurately graded the risk—or how aggressive a tumor is likely to be—[in] 82% of the tumors analyzed, while only 44% were correctly graded using a biopsy.”
Additionally, “The [AI] model also accurately predicted the disease type [in] 84% of the sarcomas tested—meaning it can effectively differentiate between leiomyosarcoma and liposarcoma—compared with radiologists who were not able to diagnose 35% of the cases,” the news release states.
“There is an urgent need to improve the diagnosis and treatment of patients with retroperitoneal sarcoma, who currently have poor outcomes,” said the study’s first author Amani Arthur, PhD, Clinical Research Fellow at The Institute of Cancer Research, London, and Registrar at The Royal Marsden NHS Foundation Trust, in the ICR news release.
“The disease is very rare—clinicians may only see one or two cases in their career—which means diagnosis can be slow. This type of sarcoma is also difficult to treat as it can grow to large sizes and, due to the tumor’s location in the abdomen, involve complex surgery,” she continued. “Through this early research, we’ve developed an innovative AI tool using imaging data that could help us more accurately and quickly identify the type and grade of retroperitoneal sarcomas than current methods. This could improve patient outcomes by helping to speed up diagnosis of the disease, and better tailor treatment by reliably identifying the risk of each patient’s disease.
“In the next phase of the study, we will test this model in clinic on patients with potential retroperitoneal sarcomas to see if it can accurately characterize their disease and measure the performance of the technology over time,” Arthur added.
Importance of Study Findings
Speed of detection is key to successful cancer diagnoses, noted Richard Davidson, Chief Executive of Sarcoma UK, a bone and soft tissue cancer charity.
“People are more likely to survive sarcoma if their cancer is diagnosed early—when treatments can be effective and before the sarcoma has spread to other parts of the body. One in six people with sarcoma cancer wait more than a year to receive an accurate diagnosis, so any research that helps patients receive better treatment, care, information and support is welcome,” he told The Guardian.
According to the World Health Organization, cancer kills about 10 million people worldwide every year. Acquisition and medical laboratory testing of tissue biopsies is both painful to patients and time consuming. Thus, a non-invasive method of diagnosing deadly cancers quickly, accurately, and early would be a boon to oncology practices worldwide and could save thousands of lives each year.
Researchers intend their new AI image retrieval tool to help pathologists locate similar case images to reference for diagnostics, research, and education
Researchers at Stanford University turned to an unusual source—the X social media platform (formerly known as Twitter)—to train an artificial intelligence (AI) system that can look at clinical laboratory pathology images and then retrieve similar images from a database. This is an indication that pathologists are increasingly collecting and storing images of representative cases in their social media accounts. They then consult those libraries when working on new cases that have unusual or unfamiliar features.
The Stanford Medicine scientists trained their AI system—known as Pathology Language and Image Pretraining (PLIP)—on the OpenPath pathology dataset, which contains more than 200,000 images paired with natural language descriptions. The researchers collected most of the data by retrieving tweets in which pathologists posted images accompanied by comments.
“It might be surprising to some folks that there is actually a lot of high-quality medical knowledge that is shared on Twitter,” said researcher James Zou, PhD, Assistant Professor of Biomedical Data Science and senior author of the study, in a Stanford Medicine SCOPE blog post, which added that “the social media platform has become a popular forum for pathologists to share interesting images—so much so that the community has widely adopted a set of 32 hashtags to identify subspecialties.”
“It’s a very active community, which is why we were able to curate hundreds of thousands of these high-quality pathology discussions from Twitter,” Zou said.
“The main application is to help human pathologists look for similar cases to reference,” James Zou, PhD (above), Assistant Professor of Biomedical Data Science, senior author of the study, and his colleagues wrote in Nature Medicine. “Our approach demonstrates that publicly shared medical information is a tremendous resource that can be harnessed to develop medical artificial intelligence for enhancing diagnosis, knowledge sharing, and education.” Leveraging pathologists’ use of social media to store case images for future reference has worked out well for the Stanford Medicine study. (Photo copyright: Stanford University.)
Retrieving Pathology Images from Tweets
“The lack of annotated publicly-available medical images is a major barrier for innovations,” the researchers wrote in Nature Medicine. “At the same time, many de-identified images and much knowledge are shared by clinicians on public forums such as medical Twitter.”
In this case, the goal “is to train a model that can understand both the visual image and the text description,” Zou said in the SCOPE blog post.
“Pathology is perhaps even more suited to Twitter than many other medical fields because for most pathologists, the bulk of our daily work revolves around the interpretation of images for the diagnosis of human disease,” wrote Jerad M. Gardner, MD, a dermatopathologist and section head of bone/soft tissue pathology at Geisinger Medical Center in Danville, Pa., in a blog post about the Pathology Hashtag Ontology project. “Twitter allows us to easily share images of amazing cases with one another, and we can also discuss new controversies, share links to the most cutting edge literature, and interact with and promote the cause of our pathology professional organizations.”
The researchers used the 32 subspecialty hashtags to retrieve English-language tweets posted from 2006 to 2022. Images in the tweets were “typically high-resolution views of cells or tissues stained with dye,” according to the SCOPE blog post.
The researchers collected a total of 232,067 tweets and 243,375 image-text pairs across the 32 subspecialties, they reported. They augmented this with 88,250 replies that received the highest number of likes and had at least one keyword from the ICD-11 codebook. The SCOPE blog post noted that the rankings by “likes” enabled the researchers to screen for high-quality replies.
They then refined the dataset by removing duplicates, retweets, non-pathology images, and tweets marked by Twitter as being “sensitive.” They also removed tweets containing question marks, as this was an indicator that the practitioner was asking a question about an image rather than providing a description, the researchers wrote in Nature Medicine.
They cleaned the text by removing hashtags, Twitter handles, HTML tags, emojis, and links to websites, the researchers noted.
The final OpenPath dataset included:
116,504 image-text pairs from Twitter posts,
59,869 from replies, and
32,041 image-text pairs scraped from the internet or obtained from the LAION dataset.
The latter is an open-source database from Germany that can be used to train text-to-image AI software such as Stable Diffusion.
Training the PLIP AI Platform
Once they had the dataset, the next step was to train the PLIP AI model. This required a technique known as contrastive learning, the researchers wrote, in which the AI learns to associate features from the images with portions of the text.
As explained in Baeldung, an online technology publication, contrastive learning is based on the idea that “it is easier for someone with no prior knowledge, like a kid, to learn new things by contrasting between similar and dissimilar things instead of learning to recognize them one by one.”
“The power of such a model is that we don’t tell it specifically what features to look for. It’s learning the relevant features by itself,” Zou said in the SCOPE blog post.
The resulting AI PLIP tool will enable “a clinician to input a new image or text description to search for similar annotated images in the database—a sort of Google Image search customized for pathologists,” SCOPE explained.
“Maybe a pathologist is looking at something that’s a bit unusual or ambiguous,” Zou told SCOPE. “They could use PLIP to retrieve similar images, then reference those cases to help them make their diagnoses.”
The Stanford University researchers continue to collect pathology images from X. “The more data you have, the more it will improve,” Zou said.
Pathologists will want to keep an eye on the Stanford Medicine research team’s progress. The PLIP AI tool may be a boon to diagnostics and improve patient outcomes and care.
Already-existing clinical laboratory blood test may be new standard for detecting Alzheimer’s biomarkers
In Sweden, an independent study of an existing blood test for Alzheimer’s disease—called ALZpath—determined that this diagnostic assay appears to be “just as good as, if not surpass, lumbar punctures and expensive brain scans at detecting signs of Alzheimer’s in the brain,” according to a report published by The Guardian.
Alzheimer’s disease is one of the worst forms of dementia and it affects more than six million people annually according to the Alzheimer’s Association. Clinical laboratory testing to diagnose the illness traditionally involves painful, invasive spinal taps and brain scans. For that reason, researchers from the University of Gothenburg in Sweden wanted to evaluate the performance of the ALZpath test when compared to these other diagnostic procedures.
Motivated to seek a less costly, less painful, Alzheimer’s biomarker for clinical laboratory testing, neuroscientist Nicholas Ashton, PhD, Assistant Professor of Neurochemistry at the University of Gothenburg, led a team of scientists that looked at other common biomarkers used to identify changes in the brain of Alzheimer’s patients. That led them to tau protein-based blood tests and specifically to the ALZpath blood test for Alzheimer’s disease developed by ALZpath, Inc., of Carlsbad, Calif.
In their JAMA article, they wrote, “the pTau217 immunoassay showed similar accuracies to cerebrospinal fluid biomarkers in identifying abnormal amyloid β (Aβ) and tau pathologies.”
In an earlier article published in medRxiv, Ashton et al wrote, “Phosphorylated tau (pTau) is a specific blood biomarker for Alzheimer’s disease (AD) pathology, with pTau217 considered to have the most utility. However, availability of pTau217 tests for research and clinical use has been limited.”
Thus, the discovery of an existing pTau217 assay (ALZpath) that is accessible and affordable is a boon to Alzheimer’s patients and to the doctors who treat them.
“The ALZpath pTau217 assay showed high diagnostic accuracy in identifying elevated amyloid (AUC, 0.92-0.96; 95%CI 0.89-0.99) and tau (AUC, 0.93-0.97; 95%CI 0.84-0.99) in the brain across all cohorts. These accuracies were significantly higher than other plasma biomarker combinations and equivalent to CSF [cerebrospinal fluid] biomarkers,” an ALZpath press release noted.
“This is an instrumental finding in blood-based biomarkers for Alzheimer’s, paving the way for the clinical use of the ALZpath pTau217 assay,” stated Henrik Zetterberg, MD, PhD (above), Professor of Neurochemistry at the University of Gothenburg and co-author of the study. “This robust assay is already used in multiple labs around the globe.” Clinical laboratories may soon be receiving doctors’ orders for pTau217 blood tests for Alzheimer’s patients. (Photo copyright: University of Gothenburg.)
Study Details
Ashton’s team conducted a cohort study that “examined data from three single-center observational cohorts.” The cohorts included:
“Participants included individuals with and without cognitive impairment grouped by amyloid and tau (AT) status using PET or CSF biomarkers. Data were analyzed from February to June 2023,” the researchers wrote.
These trials from the US, Canada, and Spain featured 786 participants and featured “either a lumbar puncture or an amyloid PET scan to identify signs of amyloid and tau proteins—hallmarks of Alzheimer’s disease,” The Guardian reported, adding that results of the University of Gothenburg’s study showed that the ALZpath pTau217 blood test “was superior to brain atrophy assessments, in identifying signs of Alzheimer’s.”
“80% of individuals could be definitively diagnosed on a blood test without any other investigation,” Ashton told The Guardian.
Diagnosis Needed to Receive Alzheimer’s Disease Treatments
“If you’re going to receive [the new drugs], you need to prove that you have amyloid in the brain,” Ashton told The Guardian. “It’s just impossible to do spinal taps and brain scans on everyone that would need it worldwide. So, this is where the blood test [has] a huge potential.”
Even countries where such drugs were not yet available (like the UK) would benefit, Ashton said, because the test, “Could potentially say that this is not Alzheimer’s disease and it could be another type of dementia, which would help to direct the patient’s management and treatment routine.”
However, Ashton himself noted the limitations of the new findings—specifically that there is no success shown yet in Alzheimer’s drugs being taken by symptom-free individuals.
“If you do have amyloid in the brain at 50 years of age, the blood test will be positive,” he said. “But what we recommend, and what the guidelines recommend with these blood tests, is that these are to help clinicians—so someone must have had some objective concern that they have Alzheimer’s disease, or [that] their memory is declining,” he told The Guardian.
Experts on the Study Findings
“Blood tests could be used to screen everyone over 50-years old every few years, in much the same way as they are now screened for high cholesterol,” David Curtis, MD, PhD, Honorary Professor in the Genetics, Evolution and Environment department at University College London, told The Guardian.
“Results from these tests could be clear enough to not require further follow-up investigations for some people living with Alzheimer’s disease, which could speed up the diagnosis pathway significantly in future,” Richard Oakley, PhD, Associate Director of Research and Innovation at the Alzheimer’s Society, UK, told The Guardian.
Though Oakley found the findings promising, he pointed out what should come next. “We still need to see more research across different communities to understand how effective these blood tests are across everyone who lives with Alzheimer’s disease,” he said.
“Expanding access to this highly accurate Alzheimer’s disease biomarker is crucial for wider evaluation and implementation of AD blood tests,” the researchers wrote in JAMA Neurology.
“ALZpath makers are in discussions with labs in the UK to launch it for clinical use this year, and one of the co-authors, Henrik Zetterberg, MD, PhD, Professor of Neurochemistry at the University of Gothenburg, is making the assay available for research use as part of the ‘biomarker factory’ at UCL,” The Guardian reported.
In the US, to be prescribed any of the available Alzheimer’s medications, a doctor must diagnose that the patient has amyloid in the brain. A pTau217 diagnostic blood test could be used to make such a diagnosis. Currently, however, the test is only available “for research studies through select partner labs,” Time reported.
“But later this month, doctors in the US will be able to order the test for use with patients. (Some laboratory-developed tests performed by certain certified labs don’t require clearance from the US Food and Drug Administration.),” Time added.
It may be that the University of Gothenburg study will encourage Alzheimer’s doctors in the UK and around the world to consider ordering pTau217 diagnostic blood tests from clinical laboratories, rather than prescribing spinal taps and brains scans for their Alzheimer’s patients.
Study findings could lead to new clinical laboratory testing biomarkers designed to assess for male infertility
Clinical laboratories are increasingly performing tests that have as their biomarkers the DNA and enzymes found in human microbiota. And microbiologists and epidemiologists know that like other environments within the human body, semen has its own microbiome. Now, a study conducted at the University of California, Los Angeles (UCLA) has found that the health of semen microbiome may be linked to male infertility.
The UCLA researchers discovered a small group of microorganisms within semen that may impair the sperm’s motility (its ability to swim) and affect fertility.
A total of 73 individuals were included in the study. About half of the subjects were fertile and already had children, while the remaining men were under consultation for fertility issues.
“These are people who have been trying to get pregnant with their partner, and they’ve been unsuccessful,” Sriram Eleswarapu, MD, PhD, a urologist at UCLA and co-author of the study, told Scientific American. “This latter group’s semen samples had a lower sperm count or motility, both of which can contribute to infertility.”
“There is much more to explore regarding the microbiome and its connection to male infertility,” said Vadim Osadchiy, MD (above), a resident in the Department of Urology at UCLA and lead author of the study, in a UCLA news release. “However, these findings provide valuable insights that can lead us in the right direction for a deeper understanding of this correlation.” Might it also lead to new biomarkers for clinical laboratory testing for male infertility? (Photo copyright: UCLA.)
Genetic Sequencing Used to Identify Bacteria in Semen Microbiome
Most of the microbes present in the semen microbiome originate in the glands of the male upper reproductive tract, including the testes, seminal vesicles and prostate, and contribute various components to semen. “Drifter” bacteria that comes from urine and the urethra can also accumulate in the fluid during ejaculation. Microbes from an individual’s blood, or his partner’s, may also aggregate in semen. It is unknown how these bacteria might affect health.
“I would assume that there are bacteria that are net beneficial, that maybe secrete certain kinds of cytokines or chemicals that improve the fertility milieu for a person, and then there are likely many that have negative side effects,” Eleswarapu told Scientific American.
The scientists used genetic sequencing to identify different bacteria species present within the semen microbiome. They found five species that were common among all the study participants. But men with more of the microbe Lactobacillus iners (L. iners) were likelier to have impaired sperm motility and experience fertility issues.
This discovery was of special interest to the team because L. iners is commonly found in the vaginal microbiome. In females, high levels of L. iners are associated with bacterial vaginosis and have been linked to infertility in women. This is the first study that found a negative association between L. iners and male fertility.
The researchers plan to investigate specific molecules and proteins contained in the bacteria to find out whether they slow down sperm in a clinical laboratory situation.
“If we can identify how they exert that influence, then we have some drug targets,” Eleswarapu noted.
Targeting Bacteria That Cause Infertility
The team also discovered that three types of bacteria found in the Pseudomonas genus were present in patients who had both normal and abnormal sperm concentrations. Patients with abnormal sperm concentrations had more Pseudomonas fluorescens and Pseudomonas stutzeri and less Pseudomonas putida in their samples.
According to the federal National Institute of Child Health and Human Development (NICHD), “one-third of infertility cases are caused by male reproductive issues, one-third by female reproductive issues, and the remaining one-third by both male and female reproductive issues or unknown factors.” Thus, learning more about how the semen microbiome may be involved in infertility could aid in the development of drugs that target specific bacteria.
“Our research aligns with evidence from smaller studies and will pave the way for future, more comprehensive investigations to unravel the complex relationship between the semen microbiome and fertility,” said urologist Vadim Osadchiy, MD, a resident in the Department of Urology at UCLA and lead author of the study, in a UCLA news release.
More research is needed. For example, it’s unclear if there are any links between the health of semen microbiome and other microbiomes that exist in the body, such as the gut microbiome, that cause infertility. Nevertheless, this research could lead to new biomarkers for clinical laboratory testing to help couples who are experiencing fertility issues.
HHS Office of Inspector General was the latest to examine the quality control problems that led to distribution of inaccurate test to clinical laboratories nationwide
Failure on the part of the Centers for Disease Control and Prevention (CDC) to produce accurate, dependable SARS-CoV-2 clinical laboratory test kits at the start of the COVID-19 pandemic continues to draw scrutiny and criticism of the actions taken by the federal agency.
In the early weeks of the COVID-19 pandemic, the CDC distributed faulty SARS-CoV-2 test kits to public health laboratories (PHLs), delaying the response to the outbreak at a critical juncture. That failure was widely publicized at the time. But within the past year, two reports have provided a more detailed look at the shortcomings that led to the snafu.
“We identified weaknesses in CDC’s COVID-19 test kit development processes and the agencywide laboratory quality processes that may have contributed to the failure of the initial COVID-19 test kits,” the OIG stated in its report.
Prior to the outbreak, the agency had internal documents that were supposed to provide guidance for how to respond to public health emergencies. However, “these documents do not address the development of a test kit,” the OIG stated.
“If the CDC can’t change, [its] importance in health in the nation will decline,” said microbiologist Jill Taylor, PhD (above), Senior Adviser for the Association of Public Health Laboratories in Washington, DC. “The coordination of public health emergency responses in the nation will be worse off.” Clinical laboratories that were blocked from developing their own SARS-CoV-2 test during the pandemic would certainly agree. (Photo copyright: Columbia University.)
Problems at the CDC’s RVD Lab
Much of the OIG’s report focused on the CDC’s Respiratory Virus Diagnostic (RVD) lab which was part of the CDC’s National Center for Immunization and Respiratory Diseases (NCIRD). The RVD lab had primary responsibility for developing, producing, and distributing the test kits. Because it was focused on research, it “was not set up to develop and manufacture test kits and therefore had no policies and procedures for developing and manufacturing test kits,” the report stated.
The RVD lab also lacked the staff and funding to handle test kit development in a public health emergency, the report stated. As a result, “the lead scientist not only managed but also participated in all test kit development processes,” the report stated. “In addition, when the initial test kit failed at some PHLs, the lead scientist was also responsible for troubleshooting and correcting the problem.”
To verify the test kit, the RVD lab needed samples of viral material from the agency’s Biotechnology Core Facility Branch (BCFB) CORE Lab, which also manufactured reagents for the kit.
“RVD Lab, which was under pressure to quickly create a test kit for the emerging health threat, insisted that CORE Lab deviate from its usual practices of segregating these two activities and fulfill orders for both reagents and viral material,” the report stated.
This increased the risk of contamination, the report said. An analysis by CDC scientists “did not determine whether a process error or contamination was at fault for the test kit failure; however, based on our interviews with CDC personnel, contamination could not be ruled out,” the report stated.
The report also cited the CDC’s lack of standardized systems for quality control and management of laboratory documents. Labs involved in test kit development used two different incompatible systems for tracking and managing documents, “resulting in staff being unable to distinguish between draft, obsolete, and current versions of laboratory procedures and forms.”
Outside Experts Weigh In
The OIG report followed an earlier review by the CDC’s Laboratory Workgroup (LW), which consists of 12 outside experts, including academics, clinical laboratory directors, state public health laboratory directors, and a science advisor from the Association of Public Health Laboratories. Members were appointed by the CDC Advisory Committee to the Director.
This group cited four major issues:
Lack of adequate planning: For the “rapid development, validation, manufacture, and distribution of a test for a novel pathogen.”
Ineffective governance: Three labs—the RVD Lab, CORE Lab, and Reagent and Diagnostic Services Branch—were involved in test kit development and manufacturing. “At no point, however, were these three laboratories brought together under unified leadership to develop the SARS-CoV-2 test,” the report stated.
Poor quality control and oversight: “Essentially, at the start of the pandemic, infectious disease clinical laboratories at CDC were not held to the same quality and regulatory standards that equivalent high-complexity public health, clinical and commercial reference laboratories in the United States are held,” the report stated.
Poor test design processes: The report noted that the test kit had three probes designed to bind to different parts of the SARS-CoV-2 nucleocapsid gene. The first two—N1 (topology) and N2 (intracellular localization)—were designed to match SARS-CoV-2 specifically, whereas the third—N3 (functions of the protein)—was designed to match all Sarbecoviruses, the family that includes SARS-CoV-2 as well as the coronavirus responsible for the 2002-2004 SARS outbreak.
The N1 probe was found to be contaminated, the group’s report stated, while the N3 probe was poorly designed. The report questioned the decision to include the N3 probe, which was not included in European tests.
Also lacking were “clearly defined pass/fail threshold criteria for test validation,” the report stated.
Advice to the CDC
Both reports made recommendations for changes at the CDC, but the LW’s were more far-reaching. For example, it advised the agency to establish a senior leader position “with major responsibility and authority for laboratories at the agency.” This individual would oversee a new Center that would “focus on clinical laboratory quality, laboratory safety, workforce training, readiness and response, and manufacturing.”
In addition, the CDC should consolidate its clinical diagnostic laboratories, the report advised, and “laboratories that follow a clinical quality management system should have separate technical staff and space from those that do not follow such a system, such as certain research laboratories.”
The report also called for collaboration with “high functioning public health laboratories, hospital and academic laboratories, and commercial reference laboratories.” For example, collaborating on test design and development “should eliminate the risk of a single point of failure for test design and validation,” the LW suggested.
CBS News reported in August that the CDC had already begun implementing some of the group’s suggestions, including agencywide quality standards and better coordination with state labs.
However, “recommendations for the agency to physically separate its clinical laboratories from its research laboratories, or to train researchers to uphold new quality standards, will be heavy lifts because they require continuous funding,” CBS News reported, citing an interview with Jim Pirkle, MD, PhD, Director, Division of Laboratory Sciences, National Center for Environmental Health, at the CDC.
