One of the world’s fastest growing medical laboratory companies in India is using digital pathology systems and AI to replace older diagnostic technologies
Artificial intelligence (AI) is gaining acceptance around the world and use of AI to analyze digital pathology images is expected to be a major disruptor to the profession of anatomic pathology. Internationally, several pathology companies already use AI-powered solutions to diagnose cancer.
One such example is Neuberg Diagnostics, a fast-growing clinical laboratory company in Chennai, India. Neuberg has been using AI to review digital pathology images for several years, according to Chairman and Managing Director GSK Velu, PhD, BPharm.
“We already use AI in our laboratories,” Velu said in an exclusive interview with Dark Daily. “Our main reference laboratories currently use digital pathology systems to support the pathologists and many of them are using AI with these digital pathology systems.
“AI and data analytics tools are being used in other departments too, such as in our wellness department where we use AI for predictive analytics,” he added. “We also use AI in our genomics division, and we are introducing AI into other divisions slowly and steadily.”
Neuberg operates 120 laboratories in an extensive network in India, South Africa, and the United Arab Emirates (UAE), and now in the US as well.
“Our idea is to enhance the access and affordability for next-generation techniques, meaning molecular diagnostics, genomics, pathology, digital pathology, proteomics, metabolomics, and all that. This is the spirit behind Neuberg Diagnostics,” said GSK Velu, PhD, BPharm (above), Chairman and Managing Director of Neuberg Diagnostics, in an exclusive interview with The Dark Report. Clinical laboratories that are considering investing in digital pathology technologies may want to follow its development at Neuberg’s Centre for Genomic Medicine in Raleigh, NC, which opened in May. (Photo copyright: Neuberg Diagnostics.)
Replacing Older Pathology Technologies
As has been happening at other anatomic pathology centers around the world, Neuberg has been using digital pathology systems to replace older technologies. “One of our largest labs is our Bangalore Reference Lab,” Velu said. “There, we do not use microscopes for histopathology, and that lab has used digital pathology for routine review of specimens for several years now.
“But because artificial intelligence is still emerging, we can’t rely on AI with all of our digital pathology systems,” he added. “Although, of course, AI is certainly an aid to everything we do with digital pathology.
“For a variety of reasons, the adaptation of artificial intelligence in anatomic pathology is not happening as effectively nor as fast as we would like,” he noted. “So, for now, we need to wait and watch a bit longer, either because adaptation by pathologists is slow, or because AI tools are still a bit of a worry for some pathologists.
Younger Pathologists Adapt Faster to Digital Pathology
One reason could be that conventional pathologists worry about relying completely on AI for any diagnosis, Velu noted. “I’m certain that the more recent generation of pathologists who are now in their 30s, and the new people coming into pathology, will start adapting more quickly to digital pathology and to AI faster than the older generation of pathologists have done.
“The younger pathologists have a greater appreciation for the potential of digital pathology, while the older pathologists don’t want to let go of conventional diagnosis methods,” he added.
“For example, we have not yet seen where pathologists are reviewing breast image scans,” he commented. “But, at the same time, AI has been well-accepted among radiologists who are reviewing breast mammography scans.”
In India and in other markets worldwide, radiologists have adapted AI tools for breast mammography scans to diagnose breast cancer, he noted. “But that’s not happening even among pathologists who are doing cancer screening,” he said.
Velu suggested that another reason for the slow adoption of AI tools in pathology is that these systems are relatively new to the market. “Maybe the AI tools that are used with digital pathology are not as reliable as we hoped they would be, or they are not fully robust at the moment,” he speculated. “That’s why I say it will take some time before the use of AI for diagnosis becomes more widespread among pathologists. So, for now, we must wait until digital pathology and AI tools work together more seamlessly.
Replacing Conventional Pathology Technologies and Methods
“When those two technologies—AI and digital pathology systems—are linked more closely, their use will take hold in a substantial way,” Velu predicted. “When that happens, they are likely to replace conventional pathology methods completely.
“Currently, we are in the early stages of a transformation,” he added. “In our labs, you can see that the transformation is ongoing. We are using digital pathology systems even in our smaller labs. Then, the staff in our smaller labs do the processing of slides to convert them to digital images and send them to our labs in the larger cities. There, the professional staff uses AI to review those digital images and issue reports based on those images.
“Using our digital pathology systems and AI in that way means that we can make that technology available even in smaller towns and villages that have access only to our smaller labs,” he commented.
Velu added that wider use of digital pathology systems could improve the quality of care that pathologists deliver to patients in a significant way, particularly in rural areas. “Here in India, we are not seeing a huge shortage of pathologists, except in rural areas and villages,” he explained. “In those places, we could run short of pathologists.
“That is the reason we are trying to adapt the use of telepathology more widely,” he noted. “To do that, we might have technicians and histologists who will do just processing of slides so that they can send the digital images to our pathologists located in larger cities. Then, those surgical pathologists will review the cases and send the reports out. That’s the model that we are trying to slowly follow here.”
As use of digital pathology images increased, many predicted that specimens would flow from the US to India. This would happen because of the belief that the lower cost of surgical pathology in India would successfully draw business away from pathology groups here in the United States.
However, Neuberg turned the tables on that belief when it announced the opening of its Neuberg Centre for Genomic Medicine (NCGM), a state-of-the-art esoteric and genetic testing laboratory in Raleigh, NC. The NCGM lab is CLIA-certified and Neuberg says it is ready to compete with labs in this country on their home turf.
These are reasons why pathologists and pathology practice administrators in the United States may want to watch how Neuberg Diagnostics continues to develop its use of digital pathology platforms and AI-powered digital image analysis tools throughout its international network of laboratories.
Because of ‘shelter in place’ orders, many anatomic pathologists are reviewing digital images from home during the COVID-19 outbreak and demonstrating the value of whole slide imaging, digital pathology, and CMS’ recent amended remote sign-out policy
COVID-19 is already triggering many permanent changes in the way healthcare is organized and delivered in the United States. However, not until the SARS-CoV-2 pandemic eases will the full extent of these changes become visible. This will be particularly true for anatomic pathology and the profession’s expanded use of telepathology, digital pathology, and whole-slide imaging.
Since early March, specimen referrals and revenues have collapsed at anatomic pathology groups and laboratories across the nation. Dark Daily’s sister publication, The Dark Report (TDR), was first to quantify the magnitude of this collapse in tissue referrals to pathology groups. In an interview with The Dark Report, Kyle Fetter, Executive Vice President and General Manager of Diagnostic Services at XIFIN, Inc., explained that pathology clients using XIFIN’s revenue cycle management services were seeing an average 40% decrease in specimens. And, for certain pathology sub-specialties, the drop-off in specimen referrals was as much as 90%. (See TDR, “From Mid-March, Labs Saw Big Drop in Revenue,” April 20, 2020.)
The College of American Pathologists (CAP) appealed to the Centers for Medicare and Medicaid Services (CMS) to allow pathologists to work remotely. In response, CMS issued a memorandum which stated, “Due to the public health emergency posed by COVID-19 and the urgent need to expand laboratory capacity, CMS is exercising its enforcement discretion to adopt a temporary policy of relaxed enforcement in connection with laboratories located at temporary testing sites under the conditions outlined herein.”
Since then, many physicians, including pathologists, have quickly adapted to working remotely in some form.
Push for Remote Pathology Services Acknowledges Anatomic Pathologist Shortage
The CMS memorandum (QSO-20-21-CLIA), which the federal agency issued to laboratory surveyors on March 26, 2020, notes that CMS will exercise “enforcement discretion to ensure pathologists may review pathology slides remotely” if certain defined conditions are met.
CMS’ decision, which “is applicable only during the COVID-19 public health emergency,” is intended to increase capacity by allowing remote site review of clinical laboratory data, results, and pathology slides.
Ordinarily, CLIA regulations for cytology (a branch of study that focuses on the biological structure of cells) state that cytology slide preparations must be evaluated on the premises of a laboratory that is certified to conduct testing in the subspecialty of cytology. However, a fast-acting Congressional letter sent by 37 members of Congress to US Department of Health and Human Services (HHS) Secretary Alex Azar II, MD, states, “it is unwise and unnecessary to overburden the remaining pathologists with excess work due to staffing shortages, thereby increasing the risk of burnout, medical error, and further shortages in staffing due to exposure. The number of COVID-19 cases will increase and peak over the next two months and will stretch existing healthcare systems to their limits.”
Decreasing Number of ‘Active Pathologists’ Drives Adoption of Telepathology, Digital Pathology, and Whole-slide Imaging
The current COVID-19 outbreak is just the latest factor in support of enabling remote review of anatomic pathology images and cases. The trend of using telepathology, whole-slide imaging (WSI), and digital pathology systems has been gathering momentum for several years. Powerful economic forces support this trend.
The Dark Report devoted its June 10, 2019, issue to a deep dive of the challenges currently facing the anatomic pathology profession. In particular, TDR noted a study published May 31, 2019, in the Journal of the American Medical Association (JAMA) Network Open, titled, “Trends in the US and Canadian Pathologist Workforces from 2007 to 2017.” The study’s authors—pathologists in the United States and Canada—reported that between 2007 and 2017 the number of active pathologists in the United States decreased from 15,568 to 12,839—a 17.53% decline.
TDR noted that these findings imply there are fewer pathologists in the United States today in active practice to handle the steady increase in the number of cases requiring diagnostic review. In turn, this situation could lead to delays in diagnoses detrimental to patient care.
In the United States, the COVID-19 pandemic created an “immediate need for remote sign-outs, reviews, and consults,” said Mike Bonham, MD, PhD (above), Chief Medical Officer for Proscia, a digital pathology software developer, in an interview with Dark Daily. “In the context of highly relevant workflow and workforce challenges, it reinforces the opportunity for wider adoption of digital pathology.” Prior to the outbreak of COVID-19, several distinct forces were driving adoption and use of digital pathology in combination with traditional microscopy, he said. (Photo copyright: Proscia.)
Distinct Forces Beginning to Reshape Anatomic Pathology
In recent years, the anatomic pathology profession has faced growing financial pressure, a shrinking workforce, and a surge in the global demand for pathology—issues that come at a time when biopsies and cancer diagnostics require greater expertise.
The UCSF School of Medicine started with frozen slide sections and moved to the broader volume of pathology slides. Since 2015, UCSF’s School of Medicine has moved toward a fully digital pathology operation and has serialized the adoption by specialty, according to Zoltan Laszik, MD, PhD, attending physician at UCSF and Professor of Clinical Pathology in UCSF’s Departments of Pathology and Laboratory Medicine.
Laszik is among a handful of specialists and digital pathology early adopters who collaborated on the new Dark Daily white paper, which is available for free download.
Through the adoption of digital pathology, glass slides are digitized using a whole-slide image scanner, then analyzed through image viewing software. Although the basic viewing functionality is not drastically different than that provided by a microscope, digitization does bring improvements in lab efficiency, diagnostic accuracy, image management, workflows, and revenue enhancements.
Additionally, artificial intelligence (AI)-based computational applications have emerged as an integral part of the digital pathology workflow in some settings, the white paper explains.
“These developments are important to anatomic pathologists because the traditional pathology business model continues to transform at a steady pace,” noted Robert L. Michel, Editor-in-Chief of The Dark Report.
Anthony Magliocco, MD, FRCPC, FCAP, President and CEO of Protean BioDiagnostics and former Professor and Chair of Pathology at Moffitt Cancer Center, is featured in the white paper as well. His new pathology service model provides routine pathology services, precision oncology, second opinions, liquid biopsies, genetics, and genomics to cancer centers from a Florida-based specialty laboratory.
To register for this important learning opportunity, click here or place this URL in your web browser: https://www.darkdaily.com/webinar/streamlined-operations-increased-revenue-higher-quality-of-care-conclusive-evidence-on-the-value-of-adopting-digital-pathology-in-your-lab/.
These digital pathology technologies represent an innovative movement shaping the present and future of pathology services. Pathologists wanting to learn more are encouraged to sign up for the May 13 webinar, which will build on the body of evidence and commentary that is included in the new white paper, and which will be available for free on-demand download following the live broadcast.
This new technology could replace needle biopsies and allow physicians to detect rejection of transplanted organs earlier, saving patients’ lives
Anatomic pathologists
may be reading fewer biopsy reports for patients with organ transplants in the
future. That’s thanks to a new technology that may be more sensitive to and
capable of detecting organ rejection earlier than traditional needle biopsies.
When clinicians can detect organ transplant rejection
earlier, patients survive longer. Unfortunately, extensive organ damage may
have already occurred by the time rejection is detected through a traditional
needle biopsy. This led a group of researchers at Emory University School of Medicine to
search for a better method for detecting organ rejection in patients with transplants.
The Emory researchers describe the method and technology
they devised in a paper published in Nature Biomedical
Engineering, titled, “Non-Invasive Early Detection of Acute Transplant
Rejection Via Nanosensors of Granzyme B Activity.” The new technology could
make it easier for clinicians to detect when a patient’s body is rejecting a
transplanted organ at an earlier time than traditional methods.
This technology also provides a running measure of processes,
so clinicians have more powerful tools for deciding on the most appropriate
dosage of immunosuppressant
drugs.
“Right now, most tests are aimed at organ dysfunction, and
sometimes they don’t signal there is a problem until organ function is below 50
percent,” Andrew
Adams, MD, PhD Co-Principal Investigator and an Associate Professor of Surgery
at Emory University School of Medicine, in a Georgia
Institute of Technology news release.
How the Technology Works
The method that Adams and his colleagues tested involves the
detection of granzyme B,
a serine protease
often found in the granules of natural killer cells
(NK cells) and cytotoxic
T cells. “Before any organ damage can happen, T cells have to produce granzyme
B, which is why this is an early detection method,” said Gabe Kwong, PhD, Assistant
Professor in the Wallace H. Coulter Department of Biomedical Engineering at
Georgia Tech and Emory University, in the news release.
The new technology is made up of sensor nanoparticles in the
shape of a ball with iron oxide in the middle. Amino acids stick out of the
ball like bristles. Each amino acid has a fluorescent molecule attached to the
tip.
The nanoparticles are injected into the patient. Their size
prevents them from gathering in the patient’s tissue or from being flushed out
through the kidneys. They are designed to accumulate in the tissue of the
transplanted organ.
If the T cells in the transplanted organ begin to produce
granzyme B, the amino acids break away from the nanoparticles, releasing the
fluorescent molecules attached to their tips. Those molecules are small enough
to be processed through the kidneys and can be detected in the patient’s urine.
Pathologists Play Crucial Role on Transplant Teams
Anatomical pathologists (histopathologists in the UK) are key
members of transplant teams for many reasons, including their ability to assess
biopsies. The current method for detecting organ transplant rejection involves
needle biopsies. It is considered the gold standard.
However, according to a paper published in the International
Journal of Organ Transplantation Medicine: “Although imaging studies
and laboratory findings are important and helpful in monitoring of the
transplanted liver, in many circumstances they are not sensitive enough. For
conditions such as rejection of the transplant, liver histology remains the
gold-standard test for the diagnosis of allograft dysfunction. Therefore,
histopathologic assessments of allograft liver
biopsies have an important role in managing patients who have undergone liver
transplantation.”
There are two main problems with needle biopsies. The first,
as mentioned above, is that they don’t always catch the rejection soon enough.
The second is that the needle may cause damage to the transplanted organ.
“The biggest risk of a biopsy is bleeding and injury to the transplanted organ,” noted Andrew Adams, MD, PhD (above), Co-Principal Investigator and an Associate Professor of Surgery at Emory University School of Medicine, in the Georgia Tech news release. “Then there’s the possibility of infection. You’re also just taking a tiny fraction of the transplanted organ to determine what’s going on with the whole organ, and you may miss rejection or misdiagnose it because the needle didn’t hit the right spot,” he added.
And, according to Kwong, even though biopsies are the gold
standard, the results represent one moment in time. “The biopsy is not
predictive. It’s a static snapshot. It’s like looking at a photo of people in
mid-jump. You don’t know if they’re on their way up or on their way down. With
a biopsy, you don’t know whether rejection is progressing or regressing.”
Future Directions of Emory’s Research
The research conducted by Adams and Kwong, et al, is in its
early stages, and the new technology they created won’t be ready to be used on patients
for some time. Nevertheless, there’s reason to be excited.
Nanoparticles are not nearly as invasive as a needle biopsy.
Thus, risk of infection or damaging the transplanted organ is much lower. And Emory’s
technology would allow for much earlier detection, as well as giving clinicians
a better way to adjust the dose of immunosuppressant drugs the patient takes.
“Adjusting the dose is very difficult but very important
because heavy immunosuppression increases occurrence of infections and patients
who receive it also get cancer more often,” said Kwong. The new technology
provides a method of measuring biological activity rates, which would give
clinicians a clearer picture of what’s happening.
The Emory team’s plan is to enhance the new sensors to
detect at least one other major cause of transplant rejection—antibodies. When
a patient’s body rejects a transplanted organ, it produces antibodies to
neutralize what it sees as a foreign entity.
“Antibodies kill their target cells through similar types of
enzymes. In the future, we envision a single sensor to detect both types of
rejection,” said Kwong.
Adams adds, “This method could be adapted to tease out
multiple problems like rejection, infection, or injury to the transplanted
organ. The treatments for all of those are different, so we could select the
proper treatment or combination of treatments and also use the test to measure
how effective treatment is.”
This line of research at Emory University demonstrates how
expanding knowledge in a variety of fields can be combined in new ways. As this
happens, medical laboratories not only get new biomarkers that can be
clinically useful without the need for invasive procedures like needle biopsies,
but these same biomarkers can guide the selection of more effective therapies.
Physicians in Saskatchewan called for changes after wait times for anatomic pathology test results reached six weeks or more
Anatomic pathologist and histopathologist shortages have plagued the single-payer healthcare systems in Canada and the United Kingdom (UK) in recent years. The consequence is increased wait times for physicians in both countries to receive medical laboratory test results, which increases wait times across the entire healthcare continuum.
However, one Canadian province significantly reduced a backlog that had pushed wait times for surgical pathology test results to six weeks or more. It did this by having its pathologists perform first-stage examinations normally completed by pathology assistants or medical technologists.
The Saskatchewan Health Authority (SHA) announced in October it had cleared nearly half of the 2,600-plus biopsies that were waiting to be processed at hospital labs in Regina and Saskatoon, the Regina Leader-Post reported.
“I think we’ve been making amazing progress in the work,” Lenore Howey, Executive Director of Laboratory Services at SHA, told the newspaper. “It’s always good to take time to know and understand your process, so that we can put the right resources in the right places.”
Getting Anatomic Pathologists Involved
Howey stated the SHA cleared cases by having pathologists “assist with the work in the first phase”—or gross examination stage—of a biopsy. This is the part of the process during which pathology assistants or medical laboratory technologists typically record the size, weight, and description of a specimen and look for pathological changes.
In addition, the SHA hired an additional pathologist assistant and three histology/cytology technologists—one on a permanent basis and two on a temporary basis. Other improvements include:
Working toward resolving problems with voice recognition transcription software being piloted in Regina for the gross examination phase of processing; and;
Implementing an electronic specimen tracking system in Saskatoon, which eventually also may be used in Regina.
Physicians Express Dissatisfaction with Wait Times
Physicians attending the Saskatchewan Medical Association’s Spring Representative Assembly in May raised the backlog issue with Health Minister Jim Reiter, complaining about the impact on patient care. At that point, the backlog of pathology cases had hit 1,662 in Regina, while Saskatoon’s caseload totaled 1,005. Many of these biopsies involve cancer patients, thus delaying a diagnosis and the start of an appropriate treatment for these patients.
“I’m trying to get things done as expeditiously as possible,” urologist Francisco Garcia, MD, told the Leader-Post, “but for the first five or six weeks, I’m handcuffed in terms of what I’m able to do.”
Now, thanks to SHA’s efforts, as of Oct. 2 specimens in progress dropped to 785 in Regina and 748 in Saskatoon. Both numbers are within range of SHA’s target of 750.
“We do not have a backlog right now,” Lenore Howey, Executive Director of Laboratory Services at SHA, told the Leader-Post. “Our system is very stable, but we do have checks and balances to put in place so that we would never get there again, which we didn’t have prior.” (Photo copyright: Saskatchewan Health Authority.)
Wait Times Impacting Patient Care Worldwide
While Saskatchewan appears to have solved its most recent pathology reporting issue, this is not the first time the province has dealt with delays in lab testing reports. In 2011, Dark Daily reported on lengthy turnaround times for anatomic pathology test reports that averaged more than 12 days, which was blamed on shortage of pathologists dating back to 2001. (See, “Pathologist Shortage and Delays in Lab Test Reports Get Publicity in Saskatchewan,” August 15, 2011.)
“Making sure pathology services can cope with current and future demand is essential if we are to ensure early diagnosis and improve outcomes for patients,” Jo Martin, PhD, President of the Royal College of Pathologists, told the BBC.
Increased workloads due to new NHS screening programs and an approaching retirement crisis—a quarter of all histopathologists in the UK are aged 55 or over—has caused the Royal College of Pathologists to call for more funded training places, better IT systems, and further investment in pathology services.
While the US healthcare system is not currently experiencing a shortage of clinical laboratory staff or anatomic pathologists, shortages in other countries illustrate the impact any delay in reporting results can have on patient care.
Computer-assisted analysis using Google’s LYNA algorithm shows significant gains in speed required to analyze stained lymph node slides and sensitivity of micrometastases detection in two recent studies
Anatomic pathologists understand the complexities of reviewing slides and samples for signs of cancer’s spread. Two studies involving a new artificial intelligence (AI) algorithm from Google (NASDAQ:GOOGL) claim their “deep learning” LYmph Node Assistant (LYNA) provides increases to both the speed at which pathologists can analyze slides and improved accuracy in detecting metastatic breast cancer within the slide samples used for the studies.
Medical laboratories and other diagnostics providers are already familiar with the improvement potential of automation and other technology-based approaches to diagnosis and analysis. Google’s LYNA is an example of how AI and machine learning improvements can serve as a supplement to—not a replacement for—the skills of experts at pathology groups and clinical laboratories.
Early research done by Google indicates that integrating LYNA into existing workflows could allow pathologists to spend less time analyzing slides for minute details. Instead, they could focus on other more challenging tasks while the AI analyzes gigapixels worth of slide data to highlight regions of concern in slides and samples for deeper manual inspection.
LYNA Achieves 99% Accuracy in Study of Metastatic Breast Cancer Detection
According to the research cited in a Google AI Blog post, roughly 25% of metastatic lymph node staging classifications would change if subjected to a second pathologic review. They further note that when faced with time constraints, detection sensitivity for small metastases on individual slides can be as low as 38%.
In findings published in Archives of Pathology and Laboratory Medicine, Google researchers analyzed whole slide images from hematoxylin-eosin-stained lymph nodes for 399 patients sourced from the Camelyon16 challenge dataset. Of those slides, researchers used 270 to train LYNA and the remaining 129 for analysis. They then compared the LYNA findings to those of an independent lab using a different scanner.
“LYNA achieved a slide-level area under the receiver operating characteristic (AUC) of 99% and a tumor-level sensitivity of 91% at one false positive per patient on the Camelyon16 evaluation dataset,” the researchers stated. “We also identified [two] ‘normal’ slides that contained micrometastases.”
Google’s algorithm later received an AUC of 99.6% on a secondary dataset.
“Artificial intelligence algorithms can exhaustively evaluate every tissue patch on a slide, achieving higher tumor-level sensitivity than, and comparable slide-level performance to, pathologists,” the researchers continued. “These techniques may improve the pathologist’s productivity and reduce the number of false negatives associated with morphologic detection of tumor cells.”
Left: sample view of a slide containing lymph nodes, with multiple artifacts: the dark zone on the left is an air bubble, the white streaks are cutting artifacts, the red hue across some regions are hemorrhagic (containing blood), the tissue is necrotic (decaying), and the processing quality was poor. Right: LYNA identifies the tumor region in the center (red), and correctly classifies the surrounding artifact-laden regions as non-tumor (blue). (Image and caption copyright: Google AI Blog.)
Faster Analysis through Software Assistance
Rapid diagnosis helps improve cancer outcomes. Yet, manually reviewing and analyzing complex digital slides is time-consuming. Time constraints might also lead to false negatives due to micrometastases or small suspicious regions that slip by pathologists undetected.
The Google research team of the study published in The American Journal of Surgical Pathology sought to gauge the impact LYNA might have on the histopathologic review of lymph nodes for trained pathologists. In their multi-reader multi-case study, researchers analyzed differences in both sensitivity of detecting micrometastases and the average review time per image using both computer-aided detection and unassisted detection for six pathologists across 70 slides.
Using the LYNA algorithm to identify and outline regions likely to contain tumors, the researchers found that sensitivity increased from 83% to 91%. The time to review slides also saw a significant reduction from 116 seconds in the unassisted mode to 61 seconds in the assisted mode—a time savings of roughly 47%.
“Although some pathologists in the unassisted mode were less sensitive than LYNA,” the researchers stated, “all pathologists performed better than the algorithm alone in regard to both sensitivity and specificity when reviewing images with assistance.”
The Future of Digital Pathology using LYNA
While the two studies show positive results, both studies also reveal shortcomings. Google highlighted both limited dataset sizes and simulated diagnostic workflows as potential concerns and areas on which to focus future studies.
Still, Google’s researchers believe that algorithms such as LYNA will help to power the future of diagnostics as healthcare in the digital era continues to mature. “We remain optimistic,” state the authors of the Google AI Blog post, “that carefully validated deep learning technologies and well-designed clinical tools can help improve both the accuracy and availability of pathologic diagnosis around the world.”
While other industries see risk in the growth of AI, both studies performed by researchers at Google show how computer-assisted workflows and machine learning could accentuate and bolster the skills of trained diagnosticians, such as anatomic pathologists and clinical laboratory technicians. By working to compensate for weak points in both human skill and computer reasoning, the outcome could be greater than either AI or humans can achieve separately.
