Editors: Raymond D. Aller, MD & Dennis Winsten
How a company’s AI targets clinical lab improvement from three angles
October 2026—Marie, Elara, and Stanley are the children of Abdul Hamid Halabi and his pathologist wife, Gazal Alsaati, MD, but not in the traditional sense. Instead, they are a suite of artificial intelligence-based apps born of the couple’s company, Clinlab.AI.
“Labs are sitting on some of the richest data in medicine and using very little of it,” says Halabi, who serves as Clinlab.AI’s CEO. “Our job is to unlock it.” This, in a nutshell, explains the company’s approach to product development, which involves embedding an understanding of a clinical laboratory’s specific patient population into its three AI copilots to better manage that lab’s quality control, operations, and billing. By parsing data from diverse sources within a lab, Clinlab.AI can help laboratories identify existing or potential problems, as well as strengthen their business overall, Halabi adds.
Armed with a background in engineering and experience working with AI in health care, Halabi first took on the challenge of developing a product to address laboratory quality control. By 2024, he had introduced Marie, named after the pioneering physicist and chemist Marie Curie, to the pathology marketplace. The virtual assistant uses more than 20 quality control algorithms, as well as Westgard rules, Six Sigma logic, and real-time patient-based QC, to monitor laboratory results and create and analyze reference ranges.
Generally, clinical labs conduct daily QC, often using manufacturer-supplied materials to test instruments and reagents, and perform a monthly Levey-Jennings procedure to further validate quality. The Clinlab.AI team recognized the value of having AI perform more intensive quality control. Therefore, Marie runs ongoing checks, including daily QC and daily Levey-Jennings charts.
In addition to conducting standard QC, Marie analyzes patient results based on past values. The program is designed to “understand” the laboratory using it, including the instruments and reagent lots the lab uses and its approach to QC. If a laboratory’s data are moving in a wonky direction, Marie flags it, Halabi says.
“Our AI looks at all the patient results and decides when it’s seeing something that doesn’t make sense,” explains Halabi. “The AI learns what that population looks like, what kinds of results they usually get, and it creates a digital twin [virtual replica] of that whole environment. When it starts seeing questionable results, it’s asking: ‘Did you add a new doctor—because those results don’t make sense?’”
In this manner, Marie creates personalized reference ranges “instead of the pathologist looking at the reference range across the United States or Germany or wherever,” says Halabi. “Now I can create a reference range for your population. If I have a lab in Dyer, Indiana, then I can localize the data for Dyer, Indiana, and the patients get treated correctly—not overtreated or undertreated.
“So we allow AI, essentially, the ability to predict what the patient results are going to be based on what it’s learned about [patients] in the past,” Halabi continues. In other words, by factoring in patient data, the system conducts quality control on the quality control.
This approach has had real-world impact. In 2024, Marie found issues with a diagnostic company’s vitamin D assay calibrator, which contributed to the calibrator being replaced.
“The company that produced that reagent produced both the calibrator and the QC,” explains Halabi. “So people calibrate the instrument; they run the QC; it passes. My patient results changed significantly despite the QC being stable. No one saw it for weeks, but the AI was saying something was wrong: ‘Your population did not change. How do you go from 30 percent vitamin D deficient to 50 percent vitamin D deficient across the board in one week?’”
As Halabi tweaked Marie, he introduced Stanley, which is designed to control pathology lab spending. (Stanley is named after Frederick Stanley Nowlan Jr., an aeronautical engineer who helped develop the concept of reliability-centered maintenance, an engineering framework that customizes maintenance strategies for individual assets.) Like Marie, Stanley creates a digital twin, but this one is a digital twin of the physicians who use a lab’s services. It is designed to extrapolate physician behaviors and other metrics to forecast daily patient volumes and reagent needs, automate purchasing, and optimize laboratory staffing levels, in part using data from Marie.
“When do physicians go on vacation? What days do they work? How many patients do they order for? What tests do they order?” Halabi says. “The AI model allowed us to predict the exact usage for each physician. Then we had the AI learn from Marie: ‘Well, Marie, tell me about the lab.’ And Marie says, ‘Well, I use this test, and this test, and this test, and these are the manufacturers I use.’
“The AI then went online and fetched all the relevant information about these tests,” Halabi continues. “How much can I get out of each box of tests? And so on. So we became very accurate at telling you how much product you need. And all the people in the lab need to do is, once every two weeks or once a month, spend an hour scanning their inventory” to determine how many items had been scanned in, how many had expired, how many are expiring soon, and how many are on hand.
Using such information, Stanley too provides cost-per-test analytics. “Laboratories have reported an average 35 percent reduction in inventory spend, elimination of reagent outages and waste, and simplified inventory management through a single monthly inventory scan,” Halabi says.
To address other clinical laboratory needs, Clinlab.AI more recently introduced Elara, an AI-driven billing assistant that reverse-engineers contracts to reveal how they work on the back end. Named after a Jupiter moon that was discovered at an observatory in Clinlab.AI’s home state of California, the software prescrubs claims for coding errors and other issues before submitting them to insurance, Halabi says.
The product automates eligible biller corrections, gives feedback on provider coding, and analyzes reimbursement trends across payers and CPT codes. It can also identify why some physicians have higher denial rates, thereby offering teaching moments. At the first laboratory to use Elara, the AI agent reduced initial denials from 30 to five percent, lowered final denials from 15 to one percent, and reduced average accounts receivable turnaround time by more than 80 percent, according to Halabi.
“The big value-add here is the ability to understand the insurance system,” says Halabi. “Obviously, it’s not one system; it’s many systems—it’s public payers, it’s private payers, and all the different permutations within those categories.”
Elara, Stanley, and Marie are offered as one platform and use HL7 standards to sit on top of a laboratory information system. They integrate with electronic medical records using HL7 or FHIR (Fast Healthcare Interoperability Resources). Laboratory managers can interact with the virtual assistants through dedicated dashboards or email alerts.
“From the perspective of independent labs, we make sure they keep their business,” says Halabi. “We make sure that, given the limited resources they have, we’re taking care of their quality, we’re reducing the cost of their supplies, and we’re making sure that they’re getting paid for the services they have.”
—Josh Baxt
CLSI releases long-awaited 2D bar-code standard
The Clinical and Laboratory Standards Institute has released the document CLSI AUTO14, “Use of 2-Dimensional Bar Coding in Clinical and Anatomic Pathology Laboratories,” after more than a decade in the making.
CLSI AUTO14 “provides context and a framework for the use of 2-dimensional (2D) bar codes in clinical and anatomic pathology laboratories, with the long-term goal of replacing linear bar codes currently in use,” according to CLSI. It is applicable to every type of laboratory specimen and asset, including blocks, slides, tubes, containers, and documents.
“The standard has engineered a robust solution that reads 2D symbols quickly and accurately on curved tubular surfaces,” says Alexis Carter, MD, formerly secretary of the CLSI document development committee. “The encoded identifier combination is globally unique and eliminates the possibility of one institution misreading a bar code as an internal accession number when the specimen is from an outside institution.” Dr. Carter is physician informaticist, Department of Pathology and Laboratory Medicine, Children’s Healthcare of Atlanta, and adjunct associate professor, Department of Pathology and Laboratory Medicine, Emory University School of Medicine.
CLSI AUTO14 is intended to educate laboratory professionals on how to specify bar-code data, print and decode bar codes, and utilize the contents of bar codes.
CLSI, 610-688-0100
Hamamatsu slide scanner systems get FDA clearance
Hamamatsu Photonics K.K. has received FDA 510(k) clearance for its Nano-Zoomer S20MD and NanoZoomer S540MD slide scanner systems.
The NanoZoomer S20MD is a compact benchtop scanner, and the NanoZoomer S540MD is designed for high-throughput laboratory operations.
“Building on the previously cleared NanoZoomer S360MD [for routine laboratory workflows], these additions broaden the availability of FDA-cleared NanoZoomer systems in the United States, giving clinical pathology laboratories greater choice when selecting a NanoZoomer system that aligns with their operational requirements,” Hamamatsu reported.
Hamamatsu, 908-231-0960
Applied BioCode and Henry Schein forge distribution deal
Applied BioCode has entered a distribution agreement with Henry Schein for its BioCode MDx-3000 system and molecular diagnostics menu.
Under the agreement, Henry Schein will distribute the MDx-3000 and its associated assays to hospitals, health systems, and reference laboratories nationwide.
The MDx-3000 is an automated, high-throughput multiplex molecular diagnostic platform for high-complexity clinical laboratories that includes upper respiratory and gastrointestinal infection panels, as well as customizable menu options.
Applied BioCode, 562-777-9800
GH Labb rebrands as Saris Health
Tampa, Fla.-based GH Labb recently changed its name to Saris Health, a move that more closely aligns the medical technology and clinical toxicology testing company’s name with its software platform.
The company’s Saris Connect proprietary ordering and reporting platform streamlines practice workflow, providing order entry in as little as eight seconds, Saris Health posted on LinkedIn.
Saris Health, 813-290-7326
Unilabs expands use of Proscia platform
Proscia has announced that Unilabs, a leading provider of medical diagnostic services in Europe, is leveraging the Concentriq platform to support digital pathology across its network.
Unilabs, which has operations in 12 countries, is already using Proscia’s Concentriq AP-Dx solution for routine diagnostic case review in several of its locations. The company expects to launch the software in additional locations by year-end.
Unilabs analyzes approximately 9 million tissue sections annually. Included in its network are locations in the United Kingdom, Denmark, Finland, Norway, Sweden, Switzerland, France, Slovakia, Czech Republic, Portugal, Spain, and Italy.
Proscia, 215-608-5411