AI-powered lung cancer screening for health systems

An AI Second Opinion for Every Lung LDCT Scan.

Categorical scoring collapses a spectrum of risk into four buckets. LungAI unfolds it back into a continuous 0 to 100 probability, catching aggressive cancers that categories miss and sparing patients the biopsies that categories overrecommend.

0.96
Nodule-level AUROC, retrospective held-out NLST dataset.
100M
Parameter Vision Transformer
CPT 0721T
Reimbursable under HOPPS, following FDA clearance
Axial CT scan showing lung nodule with LungAI overlay
2
Non-solid
Baseline
10.9 mm, 619 mm³
Malignancy Risk: 31.7
Baseline
#103/An103(F2M)
Raysum
TH 0.8 mm
WW 1500
WL -700
LungAI CADx
Sharp
Missed cancer identified

Lung-RADS categorized this nodule as 2 (benign appearance), recommending annual follow-up only. LungAI scored it 31.7 malignancy risk. Later confirmed malignant.

*Screen image is a simulated representation of LungAI output as it would appear within a third-party CADe viewer. Actual product appearance may vary. LungAI performance data from internal retrospective validation on NLST held-out data. LungAI is an investigational clinical decision support tool and does not replace physician judgment.

LungAI is an investigational medical device and has not been cleared by FDA. Oatmeal Health is preparing a 510(k) premarket notification for submission to FDA. LungAI is not commercially available in the United States. Clinical performance data reflect retrospective studies on held-out datasets; results may differ in prospective use.
Beyond Lung-RADS categories

Limitations of categorical scoring

Lung-RADS is the foundation of lung cancer screening. Published literature highlights areas where categorical scoring alone leaves room for additional quantitative context.

Reader variability
29% category disagreement

Two radiologists reading the same scan assign different Lung-RADS categories nearly a third of the time. The patient's follow-up pathway depends on which radiologist reads the scan.

29%
Inter-reader disagreement rate
False positives
36% of patients get a false alarm

More than one in three screened patients receive an abnormal result that turns out not to be cancer. Each false positive triggers unnecessary follow-up imaging, biopsies, and patient anxiety.

36%
False-positive rate (NLST substudy)
Missed cancers
Cancers diagnosed after a negative screen

Some cancers diagnosed within screening programs were not flagged on the initial LDCT. Borderline nodules in Lung-RADS 2 and 3 are where the clinical uncertainty is highest.

~15%
False-negative rate in some screening programs
How LungAI works
Zero workflow disruption

LungAI scores appear directly in your existing CADe viewer. No new workstation, no new login, no IT project. One integration with a CADe partner unlocks their entire hospital network. Compatible with GE, Philips, and Siemens scanners.

Augments Lung-RADS, does not replace it

Lung-RADS remains the basis for guideline-concordant documentation. LungAI provides an additional quantitative reference point on borderline cases. A PDF report with per-nodule score and probability curve is delivered into PACS.

Pre-FDA clearance. All performance figures are from internal retrospective validation on NLST held-out data. Results may differ in prospective clinical use. 510(k) pre-submission filed March 2026. Pivotal study at Mass General Brigham launching Q2 2026. FDA clearance target Q2 2027.

The screening network

From patient identification to diagnosis

The bottleneck in lung cancer screening is not imaging capacity. It is finding the patients who need to be screened. Two products close the gap from both ends.

Patient identification
LungIQ
Find the patients who need screening

AI analyzes EHR data to identify patients eligible for LDCT lung cancer screening. LungIQ extracts smoking status and clinical risk factors from unstructured clinical notes, then ranks every patient by their probability of meeting USPSTF criteria.

  • Surfaces candidates that manual chart review misses
  • Ranked worklist so outreach teams work top-down
  • 94% of top-ranked patients confirmed eligible
  • Works at FQHCs, health systems, and health plans
0.96
AUROC for predicting screening eligibility, validated on 35K-patient FQHC
Diagnostic support
LungAI
Score every nodule after the scan

A 0 to 100 malignancy probability for each pulmonary nodule on LDCT. LungAI's Vision Transformer architecture analyzes nodule morphology, whole-scan context, and temporal changes between current and prior exams.

  • Continuous score augments categorical Lung-RADS
  • Displays inside your existing CADe viewer
  • Models changes between current and prior scans
  • No new workstation, no workflow change
0.96
Nodule-level AUROC, retrospective held-out NLST dataset
How the consortium works
LungIQ
Identify eligible patients
Community health center runs LungIQ on its EHR. AI surfaces patients who meet screening criteria.
Imaging partner
Patient gets LDCT scan
Patient is referred to your imaging center. Standard LDCT screening exam on existing equipment.
LungAI
Radiologist reads with AI score
LungAI malignancy probability appears in the CADe viewer. Quantitative context for every nodule.

Community health centers serve the populations most at risk but most lack CT scanners. Oatmeal Health connects them with imaging partners who do. Patients reach screening. Your center gains referred scan volume. The screening gap closes.

~6%
of eligible Americans currently screened for lung cancer
ACS/JAMA, Nov 2025
~80%
of lung cancers still diagnosed at late stage, when 5-year survival falls below 10%
ACS Cancer Facts & Figures
34M+
patient lives accessible through the Oatmeal Health FQHC network

LungIQ performance validated on a 35,000-patient FQHC deployment. LungAI performance from internal retrospective validation on NLST held-out data. LungAI is not FDA-cleared and is not available for clinical use. Prospective clinical performance has not been established. 510(k) pre-submission filed March 2026.

Retrospective case examples

Where a continuous score changes the conversation

Four held-out NLST cases illustrating how a continuous malignancy probability differs from categorical Lung-RADS scoring. Retrospective research examples only.

CT scan, Case 1, Lung-RADS 2, 10.9mm nodule
Lung-RADS 2 LungAI: 31.7 Confirmed malignant
Case 1: Elevated score on a Lung-RADS 2 nodule later confirmed malignant

A 10.9mm nodule categorized Lung-RADS 2 (benign appearance, annual follow-up). In retrospective analysis on held-out NLST data, LungAI assigned a malignancy score of 31.7; the nodule was later confirmed malignant. Retrospective research example only; not a demonstration of clinical performance.

Size
10.9mm
Volume
619mm³
LungAI Score
31.7
CT scan, Case 2, Lung-RADS 3, 12.4mm nodule
Lung-RADS 3 LungAI: 77.1 Confirmed malignant
Case 2: High score on a Lung-RADS 3 nodule later confirmed malignant

A 12.4mm nodule categorized Lung-RADS 3 (probably benign, 6-month follow-up). In retrospective analysis on held-out NLST data, LungAI assigned a malignancy score of 77.1; the nodule was later confirmed malignant. Retrospective research example only; not a demonstration of clinical performance.

Size
12.4mm
Volume
998.3mm³
LungAI Score
77.1
CT scan, Case 3, Lung-RADS 4A, 13.0mm nodule
Lung-RADS 4A LungAI: 98.7 Confirmed malignant
Case 3: Very high score on a Lung-RADS 4A nodule later confirmed malignant

A 13.0mm nodule categorized Lung-RADS 4A (suspicious). In retrospective analysis on held-out NLST data, LungAI assigned a malignancy score of 98.7. Compare to Case 4, categorized 4B but assigned 0.3. The continuous scores differentiate nodules that categorical scoring treats similarly. The nodule was later confirmed malignant. Retrospective research example only; not a demonstration of clinical performance.

Size
13.0mm
Volume
1,150.3mm³
LungAI Score
98.7
CT scan, Case 4, Lung-RADS 4B, 11.2mm nodule
Lung-RADS 4B LungAI: 0.3 Benign
Case 4: Low score on a Lung-RADS 4B nodule later confirmed benign

An 11.2mm nodule categorized Lung-RADS 4B (very suspicious). In retrospective analysis on held-out NLST data, LungAI assigned a malignancy score of 0.3; the nodule was later confirmed benign. Retrospective research example only; not a demonstration of clinical performance.

Size
11.2mm
Volume
352.4mm³
LungAI Score
0.3

In this retrospective NLST sample, LungAI assigned higher probabilities to several nodules later confirmed malignant and a low probability to a nodule later confirmed benign. Retrospective research only; prospective performance has not been established. All cases shown are from a held-out subset of the NLST dataset. LungAI is not FDA-cleared and is not available for clinical use.

The stage shift argument

Earlier detection keeps patients and revenue in your system

Every stage shift from IV to I saves a life and cuts treatment costs by 4-6x. LungAI is designed to catch cancers while they are still surgical.

Stage 5-Year Survival Est. Lifetime Cost What This Means
Stage I 70-90%
~$200K Surgery + monitoring. Best outcome. LungAI target.
Stage II 50-60%
~$310K More aggressive treatment. Chemo likely.
Stage III 20-40%
~$480K Radiation + chemo + possible surgery.
Stage IV 6-10%
~$800K-$1.2M Palliative focus. Immunotherapy. 85% diagnosed here.

Illustrative estimates. For a system treating 100 lung cancer cases, shifting from Stage IV to Stage I detection saves an estimated $60M-$100M in treatment costs.

The consortium model: scan volume from community health centers

FQHCs identify LDCT-eligible patients but do not have CT scanners. They need imaging partners. Oatmeal Health connects your center with community health centers in your region who are actively identifying and referring eligible patients through LungIQ.

You get referred scan volume. FQHCs get a downstream pathway for their patients. The screening gap closes.

Reimbursement pathway

CPT 0721T is an active CMS-priced code for AI-powered quantitative image analysis. Following FDA clearance, qualifying AI-assisted reads will be billable under this code.

CPT 0721T: ~$650 per read (following FDA clearance)

510(k) pre-submission filed March 2026. Pivotal study at Mass General Brigham launching Q2 2026. HEDIS lung screening measure expected to drive payer pressure for screening volume.

From the CEO of Coreline Soft

"Our collaboration with Oatmeal Health extends beyond technology partnerships to secure insurance reimbursement-based revenue structures for actual US patients and establish leading technology in the lung cancer screening market. This demonstrates how AI diagnostic technology evolution can drive screening reimbursement structures, improved healthcare accessibility, and revenue model innovation."

KJ
Kim Jin-kook
CEO, Coreline Soft

Ready to put LungAI to work in your center?

Whether you want to see it in your CADe viewer or build a full lung screening program, the next step takes less than 30 minutes.

For radiologists
See LungAI in your workflow

A live demo showing how LungAI appears inside your CADe viewer, what the score looks like per nodule, and how the PDF report is delivered into PACS.

Request a workflow demo
For program leaders
Build a lung screening program

How Oatmeal Health drives scan referrals to your center through the consortium model, integrates LungAI into your CADe workflow, and enables CPT 0721T billing.

Talk to our team

Pre-FDA clearance. CPT 0721T billing available only following 510(k) clearance, expected 2027.