Chest X-Ray
14 findingsPneumonia, effusion, cardiomegaly, nodules and more — localised on the film, not just listed in text.
Lumen reads chest X-rays, brain MRIs, retinal fundus photos and skin lesions — and returns a prediction, a calibrated confidence score, and a Grad-CAM map highlighting the exact region that drove the call. Evidence a clinician can actually review.
Heat concentrated in the right lower lobe — consistent with the predicted finding.
Each model is paired with its own explanation layer, so the evidence travels with the prediction.
Pneumonia, effusion, cardiomegaly, nodules and more — localised on the film, not just listed in text.
Tumour classification across glioma, meningioma and pituitary, with slice-level localisation.
Grad-CAM overlays show the pixels that mattered. A confidence band tells you how sure the model is. Reviewers stay in control.
Diabetic retinopathy grading from healthy through proliferative, with vessel-level attention.
Benign vs. malignant triage on dermoscopic images.
A signed, printable PDF — findings, confidence and the overlay — in one click.
Macro-averaged across all four modalities on independent test sets.
No DICOM gymnastics, no waiting room of tabs. Drop an image and follow the read as it unfolds.
Drag in a JPG, PNG or DICOM. We validate the modality and normalise it for the right model.
The matching model runs inference and produces class probabilities in seconds.
The leading finding, severity and a calibrated risk band are surfaced together.
A Grad-CAM overlay reveals the region that drove the prediction, so you can agree or overrule.
Everything compiles into a clean, printable report ready for the patient record.
“The heatmap is the part that changed my mind. I'm not trusting a number in a box — I can see the model is looking where I'd look.”
“We use it as a second reader for retinopathy screening. Throughput is up and nothing ships without the overlay attached.”
“As a teaching tool it's unmatched — residents finally see what a model attends to versus a senior radiologist.”