Thyroid Ultrasound Images with EU-TIRADS Classification

Comprising thyroid ultrasound images classified into 5 categories based on the European EU-TIRADS scale. The classes range from EU-TIRADS 1 (no nodule) to EU-TIRADS 5 (high-risk features). It was created by yharyarias and last updated in January 2022.

Use Cases

• Train a convolutional neural network to classify thyroid ultrasound images into the 5 EU-TIRADS categories.
• Develop a model to detect high-risk features like microcalcifications or irregular margins associated with EU-TIRADS 5.
• Analyze the visual characteristics of benign (EU-TIRADS 2) versus low-risk (EU-TIRADS 3) nodules based on shape and echogenicity.
• Build a computer vision system to assist in the identification of intermediate-risk nodules (EU-TIRADS 4) characterized by mild hypoechogenicity.

Strengths

• Images are classified using the standardized 5-level European EU-TIRADS scale.
• The dataset was updated in January 2022.
• Class definitions include specific diagnostic criteria like nodule shape, margin, and echogenicity.

Limitations

• The total number of images and class distribution are unknown, which prevents assessment of sample size or potential imbalance.
• Specific image metadata, such as resolution, patient demographics, or acquisition parameters, is not detailed.
• The geographic origin of the ultrasound images is unspecified, which may limit generalizability.

Provenance

Source

yharyarias on Hugging Face.

Collection Method

Collection method for the thyroid ultrasound images is unknown.

Freshness

Last updated on 2022-01-24.