Medical Imaging AI Models (Research)

ResNet50 from the TIA Toolbox model zoo, trained on the Kather100K dataset of 100,000 hematoxylin-and-eosin colorectal histology patches. It classifies a tissue tile into one of nine categories such as tumor epithelium, stroma, lymphocytes, mucus, smooth muscle, debris, adipose, background and normal mucosa. Research use only, not a diagnostic device.

Vision Transformer (ViT) fine-tuned on chest x-ray images for multi-class thoracic findings. Given a single frontal chest radiograph it returns class probabilities across several disease categories. One of the more downloaded chest x-ray classifiers on the Hugging Face Hub. Research and education only, not a diagnostic tool.

Image classifier by prithivMLmods that labels a bone x-ray as Fractured or Not Fractured. Given a single radiograph it returns binary class scores. One of the more downloaded fracture classifiers on the Hub. Research and education only, not a diagnostic tool.

DINOv2-base backbone fine-tuned for skin-disease image classification across 31 conditions, including basal cell carcinoma, lichen planus, lupus, herpes simplex, impetigo, leprosy variants and several genodermatoses. Broader than melanoma-only models. Research and educational use only, not a diagnostic.

ResNet50 fine-tuned on the NCT-CRC-HE-45K colorectal histology dataset. It sorts an H&E tissue patch into nine classes: adipose (ADI), background (BACK), debris (DEB), lymphocytes (LYM), mucus (MUC), smooth muscle (MUS), normal mucosa (NORM), stroma (STR) and tumor epithelium (TUM). Research use only, not a diagnostic device.

ResNet18 from the TIA Toolbox zoo, trained on the PatchCamelyon (PCam) dataset of lymph-node histology patches from breast-cancer metastasis screening. It performs binary classification of a 96x96 H&E tile as tumor (metastatic tissue present) or normal. Research use only, not a diagnostic device.

Swin Transformer skin-lesion classifier trained on an ISIC-style skin cancer dataset. Predicts eight lesion classes including melanoma, basal cell carcinoma, squamous cell carcinoma, actinic keratosis, nevus, dermatofibroma, benign keratosis and vascular lesion. Research and educational use only, not for diagnosis.

Vision Transformer fine-tuned on the HAM10000 dermatoscopy dataset. Classifies a skin-lesion image into seven categories: melanoma, melanocytic nevi, basal cell carcinoma, actinic keratoses, benign keratosis-like lesions, dermatofibroma and vascular lesions. Research and educational use only, not a diagnostic tool.

ViT image classifier by dima806 that labels a facial or skin photo as dry, normal or oily skin type. Aimed at skincare and cosmetics research rather than disease detection, and not a medical diagnostic. Research and educational use only.

Patch-level colorectal pathology classifier from HistAI, built on the Hibou-L foundation model and trained on the SPIDER colorectal dataset with expert-annotated labels. It classifies a 1120x1120 H&E patch into pathology classes such as high- and low-grade adenocarcinoma, normal mucosa and other tissue types. Research use only, not a diagnostic device.

ViT base fine-tuned on brain MRI slices to classify tumor type. Given an MRI image it returns scores for glioma, meningioma, pituitary tumor or no tumor. Most-downloaded brain-tumor classifier in this search. Research and education only, not a diagnostic tool.

Vision Transformer fine-tuned on the Kaggle chest x-ray pneumonia dataset. Given a frontal chest radiograph it predicts NORMAL versus PNEUMONIA with class scores. A widely used baseline for pneumonia screening experiments. Research and education only, not a diagnostic tool.

ViT base (patch16-224, ImageNet-21k pretrained) fine-tuned on lung CT scans to flag COVID-19 findings. Takes a CT slice image and returns COVID versus non-COVID class scores. Built for research on CT-based COVID screening. Research and education only, not a diagnostic tool.

Vision Transformer fine-tuned on retinal fundus photographs to grade diabetic retinopathy severity. Given a fundus image it returns scores across the five-level scale (0 no DR through 4 proliferative DR). Most-downloaded retinopathy classifier in this search. Research and education only, not a diagnostic tool.

ViT-base classifier fine-tuned on HAM10000 dermatoscopy images with a sharpening preprocessing step. Predicts the seven standard HAM10000 lesion classes (akiec, bcc, bkl, df, mel, nv, vasc) for a single skin-lesion image. Research and educational use only, not a medical diagnostic.

Vision-transformer classifier for peripheral-blood smear images. It labels a single white-blood-cell crop as one of four leukocyte types: eosinophil, lymphocyte, monocyte or neutrophil. Trained on a public blood-cell image dataset and meant for research and teaching, not clinical hematology.