Purpose: The hippocampus, located in the medial temporal lobe (MTL), plays an essential role in learning and memory functions. Because of its frequent and early involvement in Alzheimer's disease (AD) and other neurodegenerative diseases, it is often targeted by both structural and functional imaging. Our aim is to increase the likelihood of early recognition and assessment of AD. Methods and Materials: We propose an approach that does not directly tackle the objective of hippocampal segmentation, but simply extracts from the right and left sides of a MR image two small fixed-size, parallelepiped-shaped sub images containing the hippocampi and adjacent structures (Hippocampal Boxes, HBs). We developed an automatic procedure for selecting an optimal number of HBs, starting from which we can extract both hippocampal formations in any MR image. We discriminate between controls and AD. This way MCI population can be evaluated and a prediction on the conversion to AD is made. Results: We extracted HBs from a set of 532 images from different sources, and developed a method to discriminate between converters and non converters to AD in a MCI population. The separation between Control and AD, measured as the area under the ROC curve, is 86,3%. For the Normal vs. Converted to AD cohorts, the area under the ROC curve is 88%. The forecast is checked against clinical follow up. Conclusion: The proposed approach consists in the possibility of automatically performing morphometric studies on the MTL. This procedure can quickly and reliably provide additional information in early AD diagnosis. The study has been established within the framework of the MAGIC-5 collaboration.
Automatic analysis of medial temporal lobe region for theearly assessment of Alzheimer disease (poster No. C-2464)
DE NUNZIO, Giorgio;
2010-01-01
Abstract
Purpose: The hippocampus, located in the medial temporal lobe (MTL), plays an essential role in learning and memory functions. Because of its frequent and early involvement in Alzheimer's disease (AD) and other neurodegenerative diseases, it is often targeted by both structural and functional imaging. Our aim is to increase the likelihood of early recognition and assessment of AD. Methods and Materials: We propose an approach that does not directly tackle the objective of hippocampal segmentation, but simply extracts from the right and left sides of a MR image two small fixed-size, parallelepiped-shaped sub images containing the hippocampi and adjacent structures (Hippocampal Boxes, HBs). We developed an automatic procedure for selecting an optimal number of HBs, starting from which we can extract both hippocampal formations in any MR image. We discriminate between controls and AD. This way MCI population can be evaluated and a prediction on the conversion to AD is made. Results: We extracted HBs from a set of 532 images from different sources, and developed a method to discriminate between converters and non converters to AD in a MCI population. The separation between Control and AD, measured as the area under the ROC curve, is 86,3%. For the Normal vs. Converted to AD cohorts, the area under the ROC curve is 88%. The forecast is checked against clinical follow up. Conclusion: The proposed approach consists in the possibility of automatically performing morphometric studies on the MTL. This procedure can quickly and reliably provide additional information in early AD diagnosis. The study has been established within the framework of the MAGIC-5 collaboration.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
