TY - GEN
T1 - Prediction of dementia by hippocampal shape analysis
AU - Achterberg, Hakim C.
AU - Van Der Lijn, Fedde
AU - Den Heijer, Tom
AU - Van Der Lugt, Aad
AU - Breteler, Monique M.B.
AU - Niessen, Wiro J.
AU - De Bruijne, Marleen
PY - 2010
Y1 - 2010
N2 - This work investigates the possibility of predicting future onset of dementia in subjects who are cognitively normal, using hippocampal shape and volume information extracted from MRI scans. A group of 47 subjects who were non-demented normal at the time of the MRI acquisition, but were diagnosed with dementia during a 9 year follow-up period, was selected from a large population based cohort study. 47 Age and gender matched subjects who stayed cognitively intact were selected from the same cohort study as a control group. The hippocampi were automatically segmented and all segmentations were inspected and, if necessary, manually corrected by a trained observer. From this data a statistical model of hippocampal shape was constructed, using an entropy-based particle system. This shape model provided the input for a Support Vector Machine classifier to predict dementia. Cross validation experiments showed that shape information can predict future onset of dementia in this dataset with an accuracy of 70%. By incorporating both shape and volume information into the classifier, the accuracy increased to 74%.
AB - This work investigates the possibility of predicting future onset of dementia in subjects who are cognitively normal, using hippocampal shape and volume information extracted from MRI scans. A group of 47 subjects who were non-demented normal at the time of the MRI acquisition, but were diagnosed with dementia during a 9 year follow-up period, was selected from a large population based cohort study. 47 Age and gender matched subjects who stayed cognitively intact were selected from the same cohort study as a control group. The hippocampi were automatically segmented and all segmentations were inspected and, if necessary, manually corrected by a trained observer. From this data a statistical model of hippocampal shape was constructed, using an entropy-based particle system. This shape model provided the input for a Support Vector Machine classifier to predict dementia. Cross validation experiments showed that shape information can predict future onset of dementia in this dataset with an accuracy of 70%. By incorporating both shape and volume information into the classifier, the accuracy increased to 74%.
UR - https://www.scopus.com/pages/publications/77958047198
U2 - 10.1007/978-3-642-15948-0_6
DO - 10.1007/978-3-642-15948-0_6
M3 - Conference proceeding
AN - SCOPUS:77958047198
SN - 3642159478
SN - 9783642159473
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 42
EP - 49
BT - Machine Learning in Medical Imaging - First International Workshop, MLMI 2010, Held in Conjunction with MICCAI 2010, Proceedings
T2 - 1st International Workshop on Machine Learning in Medical Imaging, MLMI 2010, Held in Conjunction with MICCAI 2010
Y2 - 20 September 2010 through 20 September 2010
ER -