Thyroid Ultrasound Images for Three Carcinoma Subtypes

900 thyroid ultrasound images are categorized into three subtypes of thyroid carcinoma: PTC, FTC, and MTC, with 300 images per class. Curated by FangDai, this dataset supports medical image classification for AI applications in thyroid cancer diagnosis.

Use Cases

• Train a multi-class image classifier to distinguish between PTC, FTC, and MTC subtypes using 900 labeled ultrasound images.
• Benchmark model performance for rare thyroid cancer detection across the three balanced carcinoma categories.
• Develop computer vision models for medical diagnosis by learning from the visual features present in thyroid ultrasound images.
• Analyze the visual characteristics differentiating Papillary, Follicular, and Medullary Thyroid Carcinoma in ultrasound imagery.

Strengths

• Contains 900 total images, providing a substantial collection for medical image analysis.
• Features a perfectly balanced distribution with 300 images each for PTC, FTC, and MTC subtypes.
• Specifically curated for AI applications in thyroid cancer diagnosis, offering a focused medical use case.

Limitations

• Limited to 900 images, which may be insufficient for training complex deep learning models without augmentation.
• Lacks metadata or clinical features beyond the image and subtype label, limiting multi-modal analysis.
• Geographic origin and patient demographics are unknown, which may introduce bias for generalizable models.

Provenance

Source

FangDai on Hugging Face.

Collection Method

Curated collection of thyroid ultrasound images.

Freshness

Last updated on 2025-12-29.