This 90-minute webinar will help pathologists and lab executives understand artificial intelligence, its many available configurations, what’s on the horizon, and how your lab can profit from it. Special offers for teams! Because so many pathologists are working remotely, Dark Daily has arranged special group rates for pathology practices. Email info@darkreport.com or call Amanda Curtis at 512-264-7103 for additional information or to register your team.
At hospitals where results of molecular COVID-19 testing can take up to several days to return, this new method for identifying potentially infected patients could improve triage
Frustrated by shortages of essential COVID-19 tests and supplies—as well by lengthy coronavirus test turn-around times—researchers at Weill Cornell Medicine have created an Artificial Intelligence (AI) algorithm that can use routine clinical laboratory test data to determine if a patient is infected with SARS-CoV-2, the coronavirus that causes the COVID-19 disease.
This is an important development because the turn-around-time (TAT) for common lab tests is generally much shorter than COVID-19 molecular diagnostics—such as real-time reverse transcription polymerase chain reaction (RT-PCR), currently the most popular coronavirus test—and certainly serological (antibody) diagnostics, which require an infection incubation time of as much as 10-14 days before testing.
Some RT-PCR diagnostic tests for COVID-19, which detect viral RNA on nasopharyngeal swab specimens, can take up to several days to return depending on the test and on the lab’s location. But routine medical laboratory tests generally return much sooner, often within minutes or hours, making this a potential game-changer for triaging infected patients.
Machine Learning Brings AI to COVID-19 Diagnostics
Advances in the use of AI in healthcare have led to the development of machine-learning algorithms that are being utilized as diagnostic tools for anatomic pathology, radiology, and for specific complex diseases, such as cancer. The Weill Cornell scientists wanted to see if alternative lab test results could be used by an algorithmic model to identify people infected with the SARS-CoV-2 coronavirus.
“When patients come to the [emergency department] and the doctor orders several panels of routine lab [tests] and also the [SARS-CoV-2] RT-PCR test, generally the routine test results come back in a couple of hours,” Sarina Yang, MD, PhD (above), one of the authors of the study, told Modern Healthcare. “So, we thought it could be useful to use the routine labs to predict whether the RT-PCR results would be positive or negative to improve the triage process.” Yang is an assistant professor in the Department of Pathology and Laboratory Medicine, and Assistant Director of the central laboratory and Director of the toxicology laboratory at Weill Cornell Medicine. (Photo copyright: Weill Cornell Medicine.)
To perform the research, the team incorporated patients’ age, sex, and race, into a machine learning model that was based on results from 27 routine lab tests chosen from a total of 685 different tests ordered for the patients. The study included 3,356 patients who were tested for SARS-CoV-2 at New York-Presbyterian Hospital/Weill Cornell Medical Center between March 11 and April 29 of this year. The patients ranged in ages from 18 to 101 with the mean age being 56.4 years. Of those patients, 1,402 were RT-PCR positive and the remaining 1,954 were RT-PCR negative.
Using a machine-learning technique known as a gradient-boosting decision tree, the algorithm identified SARS-CoV-2 infections with 76% sensitivity and 81% specificity. When looking at only emergency department (ED) patients, the model performed even better with 80% sensitivity and 83% specificity. ED patients comprised just over half (54%) of the patients used for the study.
Weill Cornell Medicine Algorithm Could Lower False Negative Test Results
The algorithm also correctly identified patients who originally tested negative for COVID-19, but who tested positive for the coronavirus upon retesting within two days. According to the researchers, these results indicated their model could potentially decrease the amount of incorrect test results.
“We are thinking that those potentially false negative patients may demonstrate a different routine lab test profile that might be more similar to those that test positive,” Fei Wang, PhD, Assistant Professor of Healthcare Policy and Research at Weill Cornell Medicine and the study’s senior author, told Modern Healthcare. “So, it offers us a chance to capture those patients who are false negatives.”
The researchers validated their model by comparing the results with patients seen at New York Presbyterian Hospital/Lower Manhattan Hospital during the same time period. Among those patients, 496 were RT-PCR positive and 968 were negative and the algorithmic model performed with 74% specificity and 76% sensitivity.
preliminarily identify high-risk SARS-CoV-2 infected patients before RT-PCR results are available,
risk stratify patients in the ED,
select patients who need relatively urgent retesting if initial RT-PCR results are negative,
help isolate infected patients earlier, and
assist in the identification of SARS-CoV-2 infected patients in areas where RT-PCR testing is unavailable due to financial or supply constraints.
Early Results of Study Promising, But More Research is Needed
Wang noted that more research is needed on the algorithm and that he and his colleagues are currently working on ways to improve the model. They are hoping to test it with different conditions and geographies.
“Our model in the paper was built on data from when New York was at its COVID peak,” he told Modern Healthcare. “At that time, we were not doing wide PCR testing, and the patients who were getting tested were pretty sick.”
At the time of the study, the positivity rate for COVID-19 at New York-Presbyterian Hospital was in the 40% to 50% range. That was substantially higher than the current positivity rate, which is in the 2% to 3% range, Modern Healthcare reported.
“This model we built in a population in New York in a certain time period, so we can’t guarantee that it will work well universally,” Wang told Modern Healthcare.
It’s exciting to think that advances in software algorithms may one day make it possible to combine routine clinical laboratory testing and create diagnostics that identify diseases in ways the individual tests were not originally designed to do.
This study is an example that researchers in AI and informatics are working to bring new tools and diagnostic capabilities to clinical laboratories. Also, this is a demonstration of how a patient’s results from multiple other types of lab tests can by analyzed using AI and similar analytical algorithms to diagnose a health condition unrelated to the original reasons for performing those tests.
If this can be demonstrated with other diseases and health conditions, it would open up one more way that pathologists and clinical laboratory scientists can contribute to more accurate diagnoses and improved selection of the most appropriate therapies for individual patients.
Working from tissue slides similar to those used by surgical pathologists, the algorithm accurately detects prostate cancer with an impressive 98% sensitivity
It could be that a new milestone has been reached on the road to using artificial intelligence (AI) to help anatomic pathologists diagnose cancer and other diseases. A research collaboration between a major American university and an Israeli company recently published a study about the ability of an AI algorithm to correctly diagnose prostate cancer.
The scientists trained the Galen Prostate AI to recognize prostate cancer by having it examine images from over a million parts of stained tissue slides taken from patient biopsies. Expert pathologists labeled each image to teach the algorithm how to distinguish between healthy and abnormal tissue. The AI was then tested on 1,600 different tissue slide images that had been collected from 100 patients seen at UPMC who were suspected of having prostate cancer.
“Humans are good at recognizing anomalies, but they have their own biases or past experience,” said Rajiv Dhir, MD, Chief Pathologist and Vice Chair of Pathology at UPMC Shadyside Hospital, Professor of Biomedical Informatics at University of Pittsburgh, and senior author of the study, in a UPMC news release. “Machines are detached from the whole story. There’s definitely an element of standardizing care.”
The image above is “of prostate cancer (represented by the heatmap) detected by the Ibex Galen Prostate [AI] solution on a biopsy that was previously diagnosed as benign by the pathologist,” stated an Ibex news release announcing the UPMC study. (Photo copyright: Ibex.)
UPMC Algorithm Goes Beyond Cancer Detection, Exceeds Human Pathologists
The researchers also noted that this is the first algorithm to extend beyond cancer detection. It reported high performance for tumor grading, sizing, and invasion of surrounding nerves—clinically important features of pathology reports.
“Algorithms like this are especially useful in lesions that are atypical,” Dhir said. “A nonspecialized person may not be able to make the correct assessment. That’s a major advantage of this kind of system.”
The algorithm also flagged six slides as potentially containing abnormal tissue that were not flagged by human pathologists. However, the researchers pointed out that this difference does not mean the AI is better than humans at detecting prostate cancer. It is probable, for example, that the pathologists simply saw enough evidence of malignancy elsewhere in the patients’ samples to recommend treatment.
Other Studies Where AI Detected Prostate Cancer
The UPMC researchers are not the first to use AI to detect prostate cancer. In February, The Lancet Oncology published a study from researchers at Radboud University Medical Center (RUMC) in the Netherlands who developed a deep learning AI system that could determine the aggressiveness of prostate cancer in certain patients.
For that research, the RUMC scientists collected 6,000 biopsies from more than 1,200 men. They then showed the biopsy images along with the original pathology reports to their AI system. Using deep learning, the AI was able to detect and grade prostate cancer according to the Gleason Grading System (aka, Gleason Score), which is used to rate prostate cancer and choose appropriate treatment options. The Gleason Score ranges from one to five and most cancers obtain a score of three or higher.
“Systems such as ours can be used in different ways. First, it can be used to screen biopsies and to filter out the easy (benign) cases. This could reduce the workload for pathologists,” said Wouter Bulten, a PhD candidate at Radboud who worked on the study, in an interview with HemOnc Today. “Second, the system can be used as a second opinion after the pathologist’s initial read. The system can flag a case if its opinion differs from that of the pathologist. It also can give feedback during the first read, showing the pathologist where to look. In this case, the pathologist needs only to confirm the opinion of the AI system.”
Can Today’s AI Outperform Human Pathologists?
In their research, the Radboud team discovered that their AI system was able to achieve pathologist-level performance and, in some cases, even performed better than human pathologists. However, they do not foresee AI replacing the need for pathologists, but rather emerging as another method to use in cancer detection and treatment.
“We see our system as an additional tool that the pathologist can use. Although our system performs very well, it still makes mistakes,” stated Bulten. “These mistakes are often different from those a human would make. We believe that when you merge the expertise of the pathologist with the second opinion of an AI system, you get the best of both worlds.”
According to the American Cancer Society, prostate cancer is the second most common cancer among men in the US, after skin cancer. The organization estimates there will be approximately 191,930 new cases of prostate cancer diagnosed and about 33,330 deaths from the disease in the US in 2020.
Though the UPMC study focused only on prostate cancer, the scientists believe their algorithm can be trained to detect other types of cancer as well. AI in clinical diagnostics is clearly progressing, however more studies will be required. Nevertheless, if AI can truly become a useful tool for anatomic pathologists to detect cancer earlier, we may see a welcomed reduction in cancer deaths.
This is another example of technology companies working to develop medical laboratory testing that consumers can use without requiring a doctor’s order for the test
Here’s new technology that could be a gamechanger in the fight against COVID-19 if further research allows it to be used in patient care. The goal of the researchers involved is to enable individuals to test for the SARS-CoV-2 coronavirus from home with the assistance of a smartphone app enhanced by artificial intelligence (AI).
Such an approach could bypass clinical laboratories by allowing potentially infected people to confirm their exposure to the coronavirus and then consult directly with healthcare providers for diagnosis and treatment.
The at-home test is being developed through a partnership between French pharmaceutical company Sanofi and San Jose, Calif.-based Luminostics, creator of a smartphone-based diagnostic platform that “can detect or measure bacteria, viruses, proteins, and hormones from swabs, saliva, urine, and blood,” according to the company’s website.
Users who wish to self-test collect a specimen from their nose via a swab and then insert that swab into a device attached to a smartphone. The device uses chemicals and nanoparticles to examine the collected sample. If the individual has the virus, the nanoparticles in the specimen glow in a way visible to smartphone cameras. The device generates data and AI in the smartphone app processes a report. The app informs the user of the results of this COVID-19 test, and it also enables the user to connect to a doctor directly through telehealth video conferencing to discuss a diagnosis.
“This partnering project could lead to another important milestone in Sanofi’s fight against COVID-19,” said Alan Main, Sanofi’s Executive Vice President, Consumer Healthcare, and Chair of the Global Self-Care Federation, in a press release. “The development of a self-testing solution with Luminostics could help provide clarity to individuals—in minutes—on whether or not they are infected.” (Photo copyright: Global Self-Care Federation.)
According to the press release, the diagnostic platform is composed of:
an iOS/Android app to instruct a user on how to run the test, capture and process data to display test results, and then to connect users with a telehealth service based on the results;
a reusable adapter compatible with most types of smartphones; and
consumables for specimen collection, preparation, and processing.
The COVID-19 test results are available within 30 minutes or less after collecting the sample, notes the Sanofi press release. Advantages cited for having a fast, over-the-counter (OTC) solution for COVID-19 testing include:
easy access and availability;
reduced contact with others, which lowers infection risk; and
timely decision-making for any necessary treatments.
The two companies plan to have their COVID-19 home-testing application available for the public before the end of the year, subject to government regulatory clearances. They intend to make their OTC solution available through consumer and retail outlets as well as ecommerce sites.
Can Sound Be Used to Diagnose COVID-19?
Another smartphone app under development records the sound of coughs to determine if an individual has contracted COVID-19. Researchers at the Swiss Federal Institute of Technology Lausanne (École Polytechnique Fédérale de Lausanne or EPFL) in Switzerland created the Cough-based COVID-19 Fast Screening Project (Coughvid), which utilizes a mobile application and AI to analyze the sound of a person’s cough to determine if it resembles that of a person infected with the SARS-CoV-2 coronavirus.
The inspiration for this project came from doctors who reported that their COVID-19 patients have a cough with a very distinctive sound that differs from other illnesses. The cough associated with COVID-19, according the EPFL website, is a dry cough that has a chirping intake of breath at the end.
“The World Health Organization (WHO) has reported that 67.7% of COVID-19 patients exhibit a ‘dry cough,’ meaning that no mucus is produced, unlike the typical ‘wet cough’ that occurs during a cold or allergies. Dry coughs can be distinguished from wet coughs by the sound they produce, which raises the question of whether the analysis of the cough sounds can give some insights about COVID-19. Such cough sounds analysis has proven successful in diagnosing respiratory conditions like pertussis [Whooping Cough], asthma, and pneumonia,” states the EPFL website.
“We have a lot of contact with medical doctors and some of them told us that they usually were able to distinguish, quite well, from the sound of the cough, if patients were probably infected,” Tomas Teijeiro Campo, PhD, Postdoc Researcher with EPFL and one of the Coughvid researchers, told Business Insider.
The Coughvid app is in its early developmental stages and the researchers behind the study are still collecting data to train their AI. To date, the scientists have gathered more than 15,000 cough samples of which 1,000 came from people who had been diagnosed with COVID-19. The app is intended to be used as a tool to help people decide whether to seek out a COVID-19 clinical laboratory test or medical treatment.
“For now, we have this nice hypothesis. There are other work groups working on more or less the same approach, so we think it has a point,” said Teijeiro Campo. “Soon we will be able to say more clearly if it’s something that’s right for the moment.”
With additional research, innovative technologies such as these could change how clinical laboratories interact with diagnosticians and patients during pandemics. And, if proven accurate and efficient, smartphone apps in the diagnosis process could become a standard, potentially altering the path of biological specimens flowing to medical laboratories.
Physicians and clinical laboratories that do business with other healthcare providers who have been denied enrollment in Medicare or had their enrollment revoked are under increased scrutiny
Efforts by the Centers for Medicare and Medicaid Services
(CMS) to crack down on fraud could soon be bolstered by artificial
intelligence (AI) tools, placing new pressure on medical
laboratories and anatomic pathology groups to ensure that their billing
practices are fully compliant with current federal “affiliations” regulations.
This is why, last October, CMS issued a Request
for Information (RFI) seeking feedback from vendors, providers, and
suppliers about the potential use of AI tools to identify cases of fraud,
waste, and abuse in billing for healthcare services. Statements from CMS
indicate that the agency plans to deepen its investigation into the affiliations
physicians and clinical laboratories have with healthcare providers that been
involved in fraudulent behavior within the Medicare program.
At present, CMS uses a variety of approaches to spot
improper claims, the RFI notes, including the use of human medical reviewers.
However, this is a costly process that allows review of less than 1% of claims.
AI tools would increase the speed and accuracy of those investigations
exponentially.
The RFI notes that AI technology could “help CMS identify
potentially problematic affiliations upon initial screening and through continuous
monitoring. One example would be a new tool or technology that would allow
easy, seamless access to state and local business ownership and registration
information that could improve CMS’ line-of-sight to potentially problematic
business relationships.”
In a blog post on the federal agency’s website, CMS Administrator Seema Verma (above) wrote that “Advanced analytics and artificial intelligence can perform rapid analysis and comparison of large-scale claims data and medical records that could allow for more expeditious, seamless and accurate medical review, and ultimately, improved payment accuracy.” [Photo copyright: Centers for Medicare and Medicaid Services.)
CMS’ New Affiliations Rule Affects Clinical Laboratories
Our sister publication, The Dark Report (TDR),
provided in-depth coverage of this rule, which allows CMS “to revoke or deny
enrollment if it finds that a provider’s or supplier’s current or previous
affiliations pose an undue risk of fraud.” (See TDR, “Labs
Must Respond to New CMS Anti-Fraud Rule,” October 14, 2019.)
“For too many years, we have played an expensive and
inefficient game of ‘whack-a-mole’ with criminals—going after them one at a
time—as they steal from our programs,” CMS Administrator Seema Verma
said in a
statement about the new rule. “These fraudsters temporarily disappear into
complex, hard-to-track webs of criminal entities, and then re-emerge under
different corporate names. These criminals engage in the same behaviors again
and again.”
As TDR reported, the rule defines four “disclosable
events” that trigger the disclosure requirements:
Uncollected debt to Medicare, Medicaid, or CHIP;
Payment suspension under a federal healthcare program;
Exclusion by the Office of Inspector General from participation in Medicare, Medicaid, or CHIP; and
Termination, revocation, or denial of Medicare, Medicaid, or CHIP enrollment.
If disclosure is required, CMS described five definitions of
an affiliation, using a five-year look-back:
Direct or indirect ownership of 5% or more in another organization;
A general or limited partnership interest, regardless of the percentage;
An interest in which an individual or entity “exercises operational or managerial control over, or directly conducts” the daily operations of another organization, “either under direct contract or through some other arrangement;”
When an individual is acting as an officer or director of a corporation; and
Any reassignment relationship.
One interesting consequence of these definitions is that
individuals or companies that invest and own an interest in a provider
organization that has one or more “disclosable events” would be flagged by the
provider at time of enrollment or re-enrollment in the Medicare program. Over
the years, some very prominent private equity companies have been investors and
owners of medical laboratory companies that owed money to Medicare or entered
into civil settlements with the federal government where the full amount of the
alleged overpayments was not recovered and the provider neither admitted nor
denied guilt. These affiliations would need to be disclosed and could be used
by CMS to deny enrollment in the Medicare program.
“Lab companies that engage in fraud and abuse—often paying illegal inducements to physicians to encourage them to order medically-unnecessary tests—distort the lab testing marketplace and capture lab test referrals that would otherwise go to compliant clinical labs and pathology groups,” stated Robert Michel, Editor-In-Chief of The Dark Report. “So, honest labs will recognize how the new rule can help suppress various types of fraud that constantly plague the clinical lab industry.” (See TDR, “Is New Medicare Affiliation Rule Good, Bad, or Ugly?” November 4, 2019.)
Other AI Applications in Healthcare
The CMS RFI also suggests other areas in which artificial
intelligence could help identify fraudulent activity, including real-time monitoring
of electronic
health records (EHR), risk
adjustment data validation (RADV) audits, and value-based payment systems.
“These tools hold the promise of more expeditious, seamless
and accurate review of chart documentation during medical review to ensure that
we are paying for what we get and getting what we pay for,” the RFI states.
“However, concerns about potential improper payments and bad actors remain. We
need to determine whether innovative new strategies, tools, and technologies
presently exist that can increase data accuracy and integrity and consequently
reduce improper payments.”
Clinical laboratories should not be surprised by any of this.
Artificial intelligence and machine learning are increasingly becoming vital
tools in today’s modern healthcare system. Nevertheless, lab leaders should
closely monitor CMS’ use of these technologies to root out fraud, as labs are
often caught up in their investigations.
