MEDICAL IP

One capability underneath everything. MEDICAL IP's own positioning is that it leads AI powered segmentation, and that is accurate as a description of the business: the segmentation and quantification models are what every product in the portfolio is built on, from the MEDIP analysis software through the DeepCatch X measurement modules to the DeepFore prediction solutions. The visible outputs, including 3D printed anatomical models and augmented reality anatomy content, are downstream renderings of a segmentation that the models produce. Remove the models and there is no remaining data asset, workflow tool or network that would retain value.

Regulatory classification does real work here in place of vendor disclosure. The segmentation and quantification products are cleared or approved as Class II devices with defined indications, and the clinician edits and accepts the segmentation as part of normal use, so a human is structurally in the loop. The higher stakes product is DeepFore, approved in Korea at Class III, which does not measure what is present but PREDICTS liver cancer occurrence and, in the Recur variant, recurrence. Prediction against a future event is a different oversight problem from measurement, and no confidence threshold, abstention behaviour or handling rule for borderline predictions is published for it. Held at B on that basis: the device pathway constrains use, but the vendor has not published how the system behaves at its own margins.

Better than most, and unevenly distributed across the portfolio. For DeepFore the method is genuinely public, because peer reviewed publication in a major journal requires describing the model, the cohort and the validation, which is a stronger form of transparency than any model card. The company also maintains a public publications page and names its products, modalities and indications specifically rather than describing an undifferentiated AI platform. What is missing is the same detail for the flagship segmentation software: no architecture, training data description or accuracy specification for MEDIP PRO was located publicly, and a buyer evaluating segmentation quality has to rely on the regulatory submission rather than anything published.

The strongest evidence record on this sourcing list and among the better ones in the index. DeepFore was published in the Journal of Hepatology, the leading journal in its field, in May 2025, then featured on the cover of the June issue and selected as a Selection of the Month in the editors' column. Cover placement and editorial selection are decisions made by the journal, not the vendor, which is exactly the kind of external judgment this index looks for. Beyond that single result: a randomised controlled study compared virtual anatomy dissection against traditional cadaver dissection and reported superior educational efficacy and satisfaction in a student subgroup; clinical results using patient specific 3D kidney models in robotic partial nephrectomy have been published; and the company has been named a Gartner reference vendor in healthcare provider research for seven consecutive years, the first Asian company to be listed. Named academic collaborators carry full credentials across Seoul National University, Seoul National University Bundang Hospital, Yonsei University College of Dentistry and Chungnam National University Hospital. The caveat that keeps this honest: MEDICAL IP is a Seoul National University Hospital spin off with its laboratory inside the university's clinical research institute, and a substantial share of the clinical testimony comes from academics at that same institution. That is normal for a university spin out and is not an accusation, but the closest validators are not independent of the company.

No product specific statement on patient data handling, retention, de identification or model training use was located. A privacy policy exists in the site footer but no substantive position on clinical image data was found. The product ingests full CT and MRI volumes and produces patient specific 3D models, and in the 3D printing business those models become physical objects leaving the institution, which raises a stewardship question that nothing published addresses.

No HIPAA statement or business associate agreement terms were located. The axis maps imperfectly to the company's primary market, where Korea's Personal Information Protection Act governs rather than HIPAA, but it maps directly to the United States operation: MEDICAL IP holds US FDA clearances, maintains a San Jose office and names Mayo Clinic, UCLA and Arizona State among its users, so HIPAA obligations plainly attach to those deployments. Nothing published addresses either regime.

A real certification sits in the stack and it is the wrong framework for this axis, which is a distinction buyers routinely miss. MEDICAL IP holds ISO 13485, obtained alongside its European CE mark in 2019. ISO 13485 is a QUALITY MANAGEMENT SYSTEM standard for medical device manufacturers, covering design control, traceability and post market surveillance. It is not an information security standard and it is not the counterpart to ISO 27001 or SOC 2, neither of which was located here, nor was any trust centre, security page or penetration test summary. Graded C rather than Not Rated because a rigorous audited quality system does bear on software engineering discipline, but a health system security review asks a different question and there is no artifact answering it. This joins the standing family of framework confusions this index tracks, alongside registered is not cleared, Breakthrough designation is not clearance, SOC 2 is an attestation not a certification, and ONC certification is neither FDA clearance nor a HIPAA attestation.

One of the broadest multi jurisdiction portfolios in the index, built steadily over a decade rather than claimed. United States FDA Class II for MEDIP PRO in November 2019 and for DeepCatch in June 2023; European CE under MDD with ISO 13485 in December 2019; UK MHRA in September 2021; China NMPA for MEDIP Pro in April 2025, stated by the company as the first for a Korean AI company; and Korean MFDS approvals spanning Class I for the ANATDEL 3D printing solution, Class II for MEDIP, DeepCatch, TiSepX and the DeepCatch X Aorta, Heart and Torso modules, and Class III for MEDIP Pneumovid, DeepFore and DeepFore Recur, the last described as a first in the liver cancer field. Korea's Ministry of Health and Welfare designated the company an Innovative Medical Device Company in January 2024. Two precisions a buyer should carry. The company writes FDA certification; FDA CLEARS Class II devices through 510(k) and does not certify them, so read the clearance rather than the wording. And the European mark was obtained under the MDD, the directive since superseded by the EU Medical Device Regulation, so confirm current MDR status before relying on European availability, the same watch this index applies to older IVDD marks in pathology.

No fairness, subgroup or population performance disclosure was located, and the gap has a specific shape here rather than a generic one. Segmentation, body composition measurement and organ volumetry are sensitive to anatomical variation across populations, and quantitative thresholds derived from one population do not transfer automatically to another. MEDICAL IP's models were developed in Korea, principally on Korean clinical data through its Seoul National University relationships, and the company now sells into the United States, China, the United Kingdom and Europe with regulatory clearance in each. No per population validation, reference range derivation or performance breakdown by ethnicity or body habitus was published. That is the question to press, and it applies with most force to the DeepCatch X measurement modules, where a normal range that is subtly wrong for a population produces systematically wrong measurements rather than visibly wrong ones.

No electronic health record integration is described and no PACS, RIS or imaging vendor is named. The product necessarily consumes DICOM imaging but no standards support, integration method or reference deployment is documented publicly. The integrations the company does publicise are graphics and infrastructure rather than clinical: MEDIP PRO connects to NVIDIA Omniverse for real time 3D collaboration, the company is an NVIDIA Inception premier partner, and it holds a letter of support from AWS. Those are compute and rendering relationships, not a path for results to reach a clinician's workflow, and a buyer should establish how output returns to the record.

No hosting model, region, tenancy or data residency commitment is published. The question is more consequential than usual because the company operates from Korea with offices in the United States and China and sells clinical imaging analysis into all three, and imaging data crossing those particular borders raises regulatory questions in every direction. The NVIDIA Omniverse and AWS relationships suggest cloud components exist but neither their scope nor their location is stated.

No price, tier or licensing model is published for any product in the portfolio, and enquiry routes through a product inquiry form. The portfolio breadth makes the gap wider than a single product vendor's would be: MEDIP PRO, the individual DeepCatch X modules, DeepFore, the ANATDEL 3D printing service and the anatomy education platform are plainly different commercial propositions with different units, and nothing indicates whether they are licensed separately, bundled, or priced per study, per seat or per site. The 3D printing business in particular is a per item service rather than software and no rate is given.

Genuinely broad across every dimension this axis measures, and evidenced rather than claimed. Modalities span CT, MRI and plain chest radiography. Clinical domains include cardiovascular, hepatic, pulmonary, infectious disease, neurosurgical and urological applications. Buyer types run from university hospitals and academic medical centres to medical and dental schools, medical device manufacturers and population health screening operations, the last evidenced by DeepCatch X supply to seventeen branches of the Korea Association of Health Promotion. Geographic reach is backed by regulatory clearance in four jurisdictions, and named users include Mayo Clinic, UCLA, Arizona State University, Seoul National University Hospital, Samsung Medical Center, Gangnam Severance Hospital and Peking Union Medical College Hospital. Few vendors in this index can evidence that range.