There are many active research projects accessing and applying shared ADNI data. Use the search above to find specific research focuses on the active ADNI investigations. This information is requested annually as a requirement for data access.
| Principal Investigator | |
| Principal Investigator's Name: | Francesca Trenta |
| Institution: | University of Catania |
| Department: | Department of Mathematics and Computer Science |
| Country: | |
| Proposed Analysis: | In the last few years, large collections of medical and neuroimages essential to training deep neural networks have gradually become a hot topic both in the research community and in the medical industry. However, the labelling process behind the creation of such large datasets is often time-consuming and require expert evaluations. Even simple tasks, such as modality classification or image classification requires a lot of efforts from expert clinicians. Hence, there is a strong need for the automatic identification of medical images, especially in the neuroscience field, to improve the ability to filter images from a vast collection of data. These considerations furnish motivation to study and develop new efficient medical image classification algorithms. Modality classification consists in identifying the type of medical images by extracting a set of discriminant visual features. Traditional methods are based on the selection of hand-made features and require prior domain knowledge. More recently, the growing interest in deep learning technologies has led to the development of several approaches to accomplish these tasks, in a much more efficient and data-driven way. Hence, we propose the usage of a Deep Learning pipeline consisting of a 3D Convolutional Neural Network (3D-CNN) suitable to perform advanced image classification. CNN is a Deep neural network composed of processing layers used to reduce an image to its key features in order to capture the spatial dependencies. Recently, CNNs have achieved outstanding performances in 2D medical image classification task. However, researchers have recently explored how to apply 3D CNN effectively in order to elaborate the volume or spatial information from sliced images. With this regard, 3D CNNs have been proposed to process medical images, considering their apparent superiority compared to 2D CNN with volumetric data in medical imaging. In fact, the main benefit of 3D CNNs is their ability to extract multi-dimension spatial features from the input volume. Using these technologies, we are aiming to improve the process of image labelling and reduce the time required by medical experts, making, in this way, the healthcare system more efficient. |
| Additional Investigators | |
| Investigator's Name: | Daniele Ravì |
| Proposed Analysis: | In the last few years, large collections of medical and neuroimages essential to training deep neural networks have gradually become a hot topic both in the research community and in the medical industry. However, the labelling process behind the creation of such large datasets is often time-consuming and require expert evaluations. Even simple tasks, such as modality classification or image classification requires a lot of efforts from expert clinicians. Hence, there is a strong need for the automatic identification of medical images, especially in the neuroscience field, to improve the ability to filter images from a vast collection of data. These considerations furnish motivation to study and develop new efficient medical image classification algorithms. Modality classification consists in identifying the type of medical images by extracting a set of discriminant visual features. Traditional methods are based on the selection of hand-made features and require prior domain knowledge. More recently, the growing interest in deep learning technologies has led to the development of several approaches to accomplish these tasks, in a much more efficient and data-driven way. Hence, we propose the usage of a Deep Learning pipeline consisting of a 3D Convolutional Neural Network (3D-CNN) suitable to perform advanced image classification. CNN is a Deep neural network composed of processing layers used to reduce an image to its key features in order to capture the spatial dependencies. Recently, CNNs have achieved outstanding performances in 2D medical image classification task. However, researchers have recently explored how to apply 3D CNN effectively in order to elaborate the volume or spatial information from sliced images. With this regard, 3D CNNs have been proposed to process medical images, considering their apparent superiority compared to 2D CNN with volumetric data in medical imaging. In fact, the main benefit of 3D CNNs is their ability to extract multi-dimension spatial features from the input volume. Using these technologies, we are aiming to improve the process of image labelling and reduce the time required by medical experts, making, in this way, the healthcare system more efficient. |
