Section: New Results

Segmentation of the hippocampus in neurodegenerative dementias

Participants : Leonardo Cruz de Souza, Marie Chupin, Maxime Bertoux, Stéphane Lehéricy, Bruno Dubois, Foudil Lamari, Isabelle Le Ber, Michel Bottlaender, Olivier Colliot [Correspondant] , Marie Sarazin.

Our team develops various applications of our automatic segmentation method SACHA to neurological disorders, in particular in neurodegenerative dementias. This research is done in close collaboration with IM2A (Institut de la Mémoire et de la Maladie d'Alzheimer, Bruno Dubois and Marie Sarazin) at Pitié-Salpêtrière hospital.

We previously showed that automatic hippocampal segmentation can discriminate patients with Alzheimer's disease (AD) from elderly control subjects, with high sensitivity and specificity. In patients with Alzheimer's disease, we further studied the relationship between hippocampal atrophy and memory deficits. We also showed that hippocampal volume loss is correlated to tau and hyperphosphorylated tau levels measured in the cerebro-spinal fluid (CSF) but not with $A{\beta }_{4}2$ levels.

Here, our objective was to study the ability of hippocampal volumetry (HV) to differentiate between two neurodegenerative dementias: Alzheimer's disease (AD) and fronto-temporal dementia (FTD). Seventy-two participants were included: 31 AD patients with predominant and progressive episodic memory deficits associated with typical AD cerebrospinal fluid (CSF) profile and/or positive amyloid imaging (PET with 11C-labeled Pittsburgh Compound B [PiB]), 26 patients with behavioral variant FTD (bvFTD) diagnosed according to consensual clinical criteria and with no AD CSF profile, and 15 healthy controls without amyloid retention on PiB-PET exam. HV were segmented with our automated method and were normalized to total intracranial volume (nHV). Significant reductions in HV were found in both AD and bvFTD patients compared with controls, but there were no significant difference between AD and bvFTD patients. Mean nHV distinguished normal controls from either AD or bvFTD with high sensitivity (80.6% and 76.9%, respectively) and specificity (93.3% for both), but it was inefficient in differentiating AD from bvFTD (9.7% specificity). There was no difference in the clinical and neuropsychological profiles according to HV in bvFTD and AD patients. In conclusion, when considered alone, measures of HV are not good markers to differentiate AD from bvFTD. Hippocampal sclerosis associated with FTD may explain the high degree of overlap in nHV between both groups.

More details in [5] .