Meta Description: Compare the top medical data annotation companies in 2027 for radiology, pathology, clinical NLP and LLM review, with clinician annotators and HIPAA controls.
Medical data annotation is the work of labeling healthcare data so machine learning models can learn from it: outlining tumors on CT and MRI slices, tagging cells on whole-slide pathology images, extracting diagnoses from clinical notes, marking instruments in surgical video, and judging whether a medical chatbot’s answer is safe. It is harder than general labeling for three reasons. The data is regulated, so every file must be de-identified and handled under HIPAA, GDPR or both. The formats are specialized, with DICOM series and gigapixel slides that ordinary tools cannot open. And many judgment calls require clinical training.
This list is for healthtech founders, medical AI and ML leads, clinical data science teams and medtech product managers choosing an outside partner for that work. We looked for vendors with a specific healthcare practice rather than a “healthcare” line on an industry page: clinician or radiologist annotators, support for medical modalities, published security credentials, and experience with the documentation regulators expect. If your project is not clinical, our list of data annotation and labeling companies covers the broader market.
Going into 2027, the market is consolidating and specializing at once. EXL completed its acquisition of iMerit in August 2026, Shaip joined Ubiquity in February 2026, and Centaur Labs now operates as Centaur.ai. Alongside them, clinician networks and radiology core labs have grown on a simple premise: as foundation models take over first-pass labeling, the scarce resource is credentialed human review, especially for generative AI outputs and FDA reference standards.
Medical Data Annotation Companies At A Glance
Company
Headquarters
Best For
Core Services
iMerit
San Jose, California
Enterprise radiology and pathology programs
Radiology and DICOM annotation, whole-slide pathology, clinical text, PHI de-identification, Ango Hub platform
Centaur.ai
Boston, Massachusetts
Large-scale expert consensus labeling
Image, text, audio, video and waveform labeling, LLM health answer evaluation, DICOM delivery
Medcase
Manhasset, New York
Matching data to a specific medical specialty
Clinician-led annotation, clinical AI validation, LLM and multimodal evaluation, expert network
Cogito Tech
Levittown, New York
Radiologist-led imaging with regulatory documentation
Radiology, pathology, cardiology, surgical video, medical coding, waveform annotation
Imaging, clinical NER and ICD coding, de-identification, physician review of model outputs
Image Core Lab
Bengaluru, India and Ardmore, Pennsylvania
Ground truth and reader studies for radiology AI and SaMD
Radiologist annotation, ground truth validation, MRMC studies, clinical trial imaging
Shaip
New York, New York
Clinical NLP and licensed healthcare datasets
Clinical entity extraction, PHI de-identification, physician dictation and medical text datasets
CloudFactory
Reading, United Kingdom
High-volume imaging work by a trained managed workforce
X-ray, CT and MRI annotation, microscopy, surgical and patient-monitoring video
CapeStart
Cambridge, Massachusetts
Medtech teams that also need regulatory writing
Medical image annotation, literature review, CER and 510(k) support, human review of AI outputs
MedAnnotate
London, United Kingdom
Smaller specialist projects across many specialties
Radiology, pathology, dermatology, dental, ophthalmology, cardiology and clinical text annotation
Best Medical Data Annotation Companies In 2027
1. iMerit — Radiologist-Led Imaging Teams And Pathologist-Led Slide Annotation
iMerit was founded in 2012 by Radha Basu and is based in San Jose, California. In August 2026, EXL completed its acquisition of the company. Healthcare is one of iMerit’s core verticals, and it is one of the few large vendors pairing its own annotation platform, Ango Hub, with in-house clinical teams.
On the imaging side, iMerit describes radiologist-led enrichment and annotation teams working on plain films, CT, MRI and ultrasound, with support for DICOM, NIfTI and NRRD formats. It offers a Radiology Annotation Suite, a DICOM annotation tool and a radiology annotation co-pilot. Pathology work on whole-slide images is handled by pathologist-led teams that annotate cells and tissue. Beyond imaging, iMerit covers clinical text and audio in medical records and trial data, and sells a PHI de-identification product. For generative AI, its Scholars network of more than 25,000 domain experts includes physicians, and the company has described a medical chatbot project in which nurses worked in a consensus workflow with escalations to a US board-certified physician.
iMerit states that it is SOC 2, ISO 27001, GDPR, HIPAA and TISAX compliant, and it says its programmatic record keeping under 21 CFR Part 11 helps clients pursue FDA 510(k) clearance. It suits enterprise medical imaging teams that want a single vendor for tooling, clinical workforce and regulatory-grade audit trails, and that are comfortable working with a supplier that is now part of a large services group.
2. Centaur.ai — 100,000+ Expert Network With A SOC 2 Type II Audit
Centaur.ai, formerly Centaur Labs, was founded in 2017 in Boston by Erik Duhaime while he was a PhD student at the MIT Center for Collective Intelligence. It announced a $16 million Series B in 2024 and has since rebranded, though some links still use the older name.
Its model is different from a traditional managed workforce. Labeling tasks are distributed to a large network of medical students, clinicians and other subject matter experts, originally through a gamified mobile app called DiagnosUs, where labelers compete and the most accurate are paid. Each case collects multiple opinions, and Centaur aggregates them, weighting contributors by measured accuracy, to produce a consensus label. The company now cites a network of more than 100,000 experts. It handles text, image, audio, video and waveform data, has worked on X-ray, CT, MRI and ultrasound as well as dermatology and retinal OCT grading, and delivers results through an API or as JSON, CSV or DICOM. It has also moved into model evaluation with CLEAR-Bench, an evaluation of LLM answers to health questions, and a Centaur Arena for expert comparison of model outputs.
Centaur states that it completes a SOC 2 Type II audit and displays a HIPAA compliance badge. It fits teams that need many independent expert reads per case, for example to build a consensus reference standard or evaluate a medical LLM at scale.
3. Medcase — 20,000+ Credentialed Clinicians Across 100+ Specialties
Medcase describes itself as a HIPAA-compliant medical data annotation, expert network and clinical product validation company. A 2021 funding announcement dated its founding to 2017 and described it as a Boston and Israel based marketplace for on-demand clinicians; today the company lists a New York address in Manhasset and says it operates fully remote. Niv Shochat is chief executive.
The core of the business is its clinician network. Medcase says it has more than 20,000 licensed, credentialed clinicians in over 140 countries, including more than 10,000 in the United States, spanning 100+ specialties and 200+ subspecialties. Rather than training generalists, it matches each project to clinicians by specialty, subspecialty, fellowship, board certification, geography, language and case volume, and states that every label is placed by a licensed healthcare professional matched to the data. It annotates medical imaging, clinical text, EHR data, video, waveform and audio, and runs a separate clinical AI validation service that includes LLM and multimodal evaluation, which it says is built around FDA and EU AI Act requirements. Recruitment, credentialing, onboarding, training, scheduling, payments and QA are all managed by Medcase.
The company cites HIPAA and ISO 27001 compliance. Medcase is best for projects where the reader’s specialty matters more than raw volume, such as an ophthalmology validation study or a physician panel grading a clinical summarization model, and for teams that already have tooling and need qualified clinicians on demand.
4. Cogito Tech — Board-Certified Radiologists And 21 CFR Part 11 Workflows
Cogito Tech is a data labeling company headquartered in Levittown, New York, with roots it dates back to 2011. Its medical practice is unusually detailed, with service lines for digital radiology, pathology, cardiac imaging, dental, ophthalmology, surgical AI, clinical data tagging and medical generative AI.
The company says its imaging work is led by board-certified radiologists and other medical professionals, supported by specialized radiology annotators. Its digital pathology work is described as guided by pathologists, its cardiology projects involve cardiologists from hospital networks, and its clinical text work is done by medical lexicon experts. Modalities include X-ray, CT, MRI, ultrasound and PET, plus ECG and EEG waveforms and video of surgical and minimally invasive procedures. Rather than forcing clients onto a proprietary tool, Cogito works inside platforms such as RedBrick, V7 and Labelbox. The company says it has more than eight years of expertise in medical data and AI.
Cogito displays ISO 27001, ISO 9001, SOC 2, HIPAA, GDPR and CCPA credentials, refers to 21 CFR Part 11 in its medical work, and maintains a dedicated FDA compliance and support page. It also offers DataSum, a framework it positions as a way to document the traceability of training data. Cogito suits medtech and diagnostics companies that want radiologist-led labels inside their existing annotation platform, along with the documentation trail that a regulatory submission will need.
5. Welo Data — Multilingual Clinical Training Data Across 155+ Locales
Welo Data is the AI training data and evaluation business of Welocalize, the New York headquartered language services company. It works as a managed partner rather than a self-serve marketplace and runs a dedicated healthcare program.
Its healthcare offering spans medical imaging annotation, clinical named entity recognition and ICD coding, patient-facing conversational AI, reinforcement learning from medical feedback, patient data de-identification, and clinical trial and real-world evidence annotation. On imaging, it covers DICOM CT, MRI and X-ray, pathology slides, retinal imaging, ultrasound video and surgical footage. On text and audio, it covers EHRs, discharge summaries, clinical notes, referral letters, prescription data, telehealth recordings and dictation. The workforce combines annotators with medical imaging backgrounds, clinical documentation specialists, nursing professionals and life sciences researchers, and Welo says physician and specialist reviewers evaluate model outputs in its RLHF work. Its wider network lists more than 500,000 expert evaluators across 300+ domains and 155+ locales.
On credentials, Welo’s healthcare page lists ISO 13485:2016, the quality management standard for medical devices, alongside ISO/IEC 27001, ISO/IEC 27701, ISO 9001, SOC 2 Type II and GDPR compliance. Buyers who need ISO 13485 should request the current certificate during vendor qualification. Welo is a good fit for medtech companies whose own quality systems require suppliers with medical device certification, and for teams building multilingual clinical NLP or patient-facing assistants that need native-language clinical reviewers.
6. Image Core Lab — Board-Certified Radiologists For Ground Truth And MRMC Studies
Image Core Lab is an imaging core lab with offices in Whitefield, Bengaluru, India and Ardmore, Pennsylvania. Its primary business is imaging support for multi-center clinical trials run by biopharma companies, CROs and device makers, and it has extended that clinical infrastructure into radiology AI and software as a medical device (SaMD) development.
That origin sets it apart from labeling vendors. Its reads are performed by fellowship-trained radiologists board-certified in the United States, the United Kingdom and India, and its AI work is framed around the steps that regulators scrutinize most: data annotation, ground truth validation and multi-reader multi-case (MRMC) studies, the reader study design commonly used to show that a radiology AI tool improves or matches clinician performance. For device companies, it describes its role as helping to build and validate radiology AI and SaMD products from annotation through to validation.
Image Core Lab lists HIPAA compliance and ICH-GCP alignment, and says its platforms are certified under 21 CFR Part 11, the FDA rule on electronic records and signatures. It does not publish a radiologist headcount or modality list on its site, so buyers should confirm reader numbers and subspecialty coverage during scoping. It is best suited to radiology AI teams approaching a regulatory submission who need a clinical-trial-grade reference standard and reader study, rather than bulk training labels.
7. Shaip — 30 Million Patient Notes For Clinical NLP
Shaip has been operational since 2019, when, by its own account, a project for a Fortune 10 company led to its creation. In February 2026 it joined Ubiquity Global Services, a New York headquartered customer experience company, and now operates as a specialized AI data platform within that group, with primary operations in Ahmedabad, India. Shaip lists more than 150 team members and 10,000+ in-house experts.
Its healthcare strength is clinical language rather than imaging. Shaip grew out of medical transcription work, and its healthcare catalog lists 30 million patient notes and 250,000 hours of audio, including physician dictation, patient and doctor conversations, and longitudinal patient records, along with MRI images. Its services include human-powered extraction of clinical entities from unstructured records to train NLP models, and an API for de-identifying protected health information that removes direct identifiers. One healthcare case study describes delivering 10,000 de-identified, labeled records, and Shaip says it fine-tuned a model on 30 million real patient charts.
Shaip states that projects run on its patented Shaip Data Platform under GDPR, HIPAA, SOC 2 Type II, ISO 9001 and ISO 27001. It does not publish a clinician annotator count, so teams needing physician-level labeling should ask how reviewers are credentialed. Shaip suits teams building clinical NLP, ambient scribing or coding models who want licensed, de-identified medical text and audio alongside annotation.
8. CloudFactory — ISO 27001 Managed Workforce With HIPAA Business Associate Validation
CloudFactory was founded in Nepal in 2010 and is now headquartered in Reading, United Kingdom. It runs a large managed workforce for AI data work and has a long-running healthcare practice covering medical imaging, medical research and medical devices.
The company takes a clear position on who should do the labeling. In its published guidance, CloudFactory argues that radiologists and other specialists rarely have the bandwidth for bulk annotation, and that medical professionals should instead train and guide annotation teams through courses, one-to-one sessions and shared best practice. Its own trained medical image data analysts pass multi-level assessments and are monitored on quality. The work covers X-ray, MRI and CT, microscopy, tissue and cell imaging, and video for surgical robotics and patient monitoring, in client tools or any major 2D and 3D platform. In one microscopy project, its analysts annotated 1.6 million individual cells across 5,000 images, and in a patient fall-prevention project, 72,000 video images.
CloudFactory holds ISO/IEC 27001:2022 certification, has been validated against the SOC 2 Trust Service Criteria for security, availability and confidentiality, and received a third-party HIPAA compliant business associate attestation in 2022. It fits teams that have clinicians to set protocols and review edge cases but need a large, well-secured workforce for volume work at lower cost than physician reads.
9. CapeStart — Medical Image Annotation Plus 510(k) And CER Support
CapeStart is headquartered in Cambridge, Massachusetts, with delivery centers in Nagercoil and Chennai, India. It reports more than 600 AI engineers, data scientists and domain experts and over 13 years of delivering AI and software work, with a main focus on regulated life sciences.
Its annotation offering sits inside a wider life sciences practice, which is the reason to consider it. On the data side, CapeStart provides specialized medical image annotation, human-in-the-loop labeling and supervised data preparation for model development, plus ProNotate, a human review, quality check and validation service for AI outputs and datasets. On the regulatory side, its medtech team offers submission support for clinical evaluation reports (CER) and FDA 510(k) filings, post-market surveillance and evidence generation aligned to EU MDR and FDA expectations. It also runs MadeAi, its life sciences evidence platform for literature review, pharmacovigilance, HEOR, medical affairs and regulatory work.
CapeStart does not name specific security certifications or clinician annotator credentials on its site, so buyers should request its security documentation and the qualifications of the team assigned to imaging work. Its value is the combination: training data, output review, literature review and regulatory writing for the same device from one partner, which suits smaller device makers without a large regulatory affairs team.
10. MedAnnotate — Clinician-Led Annotation With FDA 510(k) Pathology Experience
MedAnnotate is a clinician-led medical data annotation service operated by MedSolutionX, a London based holding company for specialist healthcare businesses. It is the youngest and smallest company on this list, citing two years of clinical annotation work and more than 50 healthcare consultants and subject matter experts, and it describes itself as an expert service rather than a software platform.
Its specialty coverage is broad for its size. MedAnnotate works across radiology (X-ray, CT, MRI and ultrasound), digital pathology (whole-slide images, H&E, IHC and cytology), dermatology, dental imaging (OPG, CBCT and intraoral), surgical and endoscopic video, ophthalmology (fundus and OCT), cardiology (echo, angiography and ECG), and clinical text such as radiology reports, discharge summaries and EHR notes for NER, relation extraction and ICD-10 coding. Its FAQ says radiologists, pathologists, dentists and dermatologists are involved, with clinically trained annotators supervised by medical consultants. The workflow ends with a senior medical consultant reviewing all annotations for diagnostic correctness, and delivered records include inter-annotator agreement scores and revision history.
MedAnnotate says its pathology team has direct experience supporting datasets for FDA 510(k) and De Novo submissions, and that it can align protocols and documentation to regulatory requirements. It describes its workflows as HIPAA-aligned rather than claiming a certification. It suits startups and research groups with focused, specialty-specific projects that want clinician involvement and direct access to the reviewing consultants without enterprise minimums.
How To Choose A Medical Data Annotation Company
Who Is Actually Doing The Labeling?
Ask for the qualifications of the people who will touch your data, not the company’s overall network size. There are three common models on this list: credentialed clinicians placing every label (Medcase, MedAnnotate, Image Core Lab), trained annotators working under clinician-written protocols with specialist review (CloudFactory, iMerit, Cogito Tech), and large expert crowds whose reads are combined into a consensus (Centaur.ai). Each can work. The right one depends on the task: segmenting a liver on CT is teachable, while grading a borderline dysplasia on a pathology slide is not. Get named credentials, the ratio of clinician reviewers to annotators, and how reviewers are matched to subspecialties.
How Is Patient Data De-Identified And Protected?
Any vendor handling protected health information should sign a business associate agreement under HIPAA and, for European data, a data processing agreement under GDPR. Ask whether de-identification happens before data leaves your environment or on the vendor’s side, how burned-in text on images and DICOM header fields are scrubbed, and how free-text notes are handled. Then check credentials against claims: an ISO 27001 certificate or a SOC 2 Type II report is an independent audit, while phrases such as “HIPAA-aligned” or “ISO-aligned” describe intent.
Can The Vendor Handle Your Modalities And File Formats?
Medical data is not a folder of JPEGs. Confirm native support for DICOM series and multiplanar 3D viewing for CT and MRI, NIfTI or NRRD for research pipelines, gigapixel whole-slide images for pathology, multi-frame ultrasound and surgical video, and waveforms such as ECG and EEG. Ask whether the vendor works in your existing platform, how 16-bit image data is preserved, and what export formats you will receive.
What Does The Quality And Adjudication Workflow Look Like?
Medical labels are frequently disputed even among experts, so a single read is rarely enough for a reference standard. Ask how many independent reads each case receives, how disagreements are resolved (senior reader adjudication, majority vote, or weighted consensus), and whether you will receive inter-annotator agreement metrics alongside the labels. For generative AI evaluation, ask how physician graders are calibrated against each other and how safety-critical errors are escalated. Good vendors will share sample QA reports before you sign.
Is The Work Ready For A Regulatory Submission?
If your model is headed for FDA clearance or a CE mark, the provenance of your training and test data will be reviewed. Ask whether the vendor keeps audit trails and electronic records consistent with 21 CFR Part 11, whether its quality system is certified to ISO 13485, how annotator credentials and training are documented, and whether it can run or support reader studies such as MRMC designs. Separating training data vendors from the team that builds your independent test set is also worth considering, since regulators look closely at independence between the two.
Conclusion
The right medical data annotation partner depends on what the labels are for. Teams that need volume across imaging and text with enterprise security will look first at iMerit, Cogito Tech, CloudFactory and Shaip. Teams that need specific clinical expertise or many independent expert reads should compare Medcase, Centaur.ai and MedAnnotate. Device makers approaching a submission will find the most relevant regulatory infrastructure at Image Core Lab, Welo Data and CapeStart.
Whichever direction you take, run a paid pilot on your hardest cases, ask for the credentials of the people who will label your data, and check certifications against current certificates or audit reports rather than website badges. In a consolidating market, it is also worth confirming ownership and roadmap at contract time.
