Dataset for DTViT

Home » Dataset Download » Dataset for DTViT

Dataset for DTViT

Datasets

Dataset for DTViT

File

Dataset for DTViT

Use Case

Dataset for DTViT

Description

The Intracerebral Hemorrhage (ICH) dataset provides real-world CT images for the classification of ICH and normal brain conditions, as well as three ICH types based on hemorrhage location: Deep, Subcortical, and Lobar. To enhance diagnosis, we introduce the Dual-Task Vision Transformer (DTViT), which uses Vision Transformer (ViT) for feature extraction and multilayer perceptron (MLP) decoders for simultaneous ICH detection and classification.

Description:

Intracerebral hemorrhage (ICH) is a critical and abrupt medical condition resulting from the rupture of blood vessels in the brain, causing permanent damage to brain tissue and often leading to significant functional disabilities or death. Diagnosing and analyzing ICH primarily depend on brain CT imaging. Given the urgency of ICH, swift and accurate analysis of CT images is essential for devising effective treatment plans. However, the complexity of ICH CT images and the shortage of specialized radiologists present considerable challenges.

Download Dataset

To overcome these challenges, we have assembled a real-world dataset for classifying both ICH and normal brain CT images. This dataset also supports the classification of three ICH types based on hemorrhage location: Deep, Subcortical, and Lobar.

Alongside the dataset, we introduce a novel neural network architecture, the Dual-Task Vision Transformer (DTViT), designed for automated ICH image classification and diagnosis. The DTViT leverages the Vision Transformer (ViT) encoder and its attention mechanisms to extract features from CT images. Our framework integrates two multilayer perceptron (MLP)-based decoders to concurrently detect the presence of ICH and classify the three hemorrhage types.