Dataset Introduction:
Cardiac ultrasound imaging, or echocardiography, is a crucial diagnostic tool for cardiologists, enabling them to examine the internal structure and function of the heart. It provides essential insights into the shape of the heart chambers and the performance of the heart valves. Echocardiography is key in diagnosing various heart conditions such as heart failure, valvular diseases, and congenital heart defects, among others.
The Cardiac Assessment and Classification of Ultrasound (CACTUS) dataset is a comprehensive collection of graded ultrasound images obtained from scans of a CAE Blue Phantom, covering different heart views and quality levels. The dataset includes 37,736 cardiac ultrasound images, each classified and graded on a scale from 0 to 10.
Ultrasound Parameters Tuning:
The image collection was conducted using various ultrasound parameter settings. The table below provides an overview of the ultrasound parameter tuning and their respective value ranges.
Cardiac Ultrasound Images:
The dataset contains five primary cardiac views: apical four-chamber (A4C), subcostal four-chamber (SC), parasternal long axis (PL), and two parasternal short axis views—one for the aortic valve (PSAV) and one for the mitral valve (PSMV). A sixth class consists of random ultrasound images captured from different probe positions.
– Number of A4C images: 7422
– Number of SC images: 6345
– Number of PL images: 6102
– Number of PSAV images: 5832
– Number of PSMV images: 6014
– Number of Random images: 6021
Image Grading:
The CACTUS dataset is evaluated by imaging experts who developed a grading system based on two main factors: completeness and clarity. Completeness refers to how much of the targeted cardiac structure is visible, with higher grades given to images displaying the entire structure. Clarity assesses the image’s brightness and its clarity, ensuring it is free from speckles and noise.
Dataset for Deep Learning Development:
This dataset was used to develop an AI framework for classifying and grading cardiac ultrasound images. The dataset folder includes real-time scan videos linked to the AI framework, showcasing the results produced by the system. Additionally, it contains videos for further deep learning training, along with associated grading and classification files.
Dataset Link: Download the CACTUS database
Paper: H. Elmekki, A. Alagha, H. Sami, A. Spilkin et al. CACTUS : An open Dataset and Framework for Automated Cardiac Assessment and Classification of Ultrasound Images Using Deep Transfer Learning. Under review (Computers in Biology and Medicine)