Cerebriu’s clinical solutions are built on research and development within AI and deep learning based image quantification. Explore our core feature timeline below.
Whole brain, ventricle, and hippocampal atrophy measurements are important biomarkers for especially Alzheimer’s disease. Our ML algorithms Radius offers single MRI, T1 atrophy quantification through brain segmentation and reference populations (see Brain segmentation in the 2018 timeline entry).
Radius, however, also offers atrophy rate quantification using advanced registration, simultaneous bias correction, and accurate volume difference calculations. The is important for clinical monitoring of patient disease progression but also plays a significant role in efficacy quantification for clinical drug trials.
Hippocampal Texture – early detection of Alzheimer’s disease using MR hippocampal texture.
White matter lesions
Segmentation and quantification of white matter lesions from brain MRI is an important driver to understand vascular pathology in many neurological disorders. White matter lesions or white matter hypo-/hyperintensities are typically quantified from T1 or FLAIR sequences respectively.
Our ML algorithms supports automatic quantification of white matter lesions from both T1 and FLAIR. Additionally, a novel Cerebriu approach improves T1-based quantification by simultaneous and coupled imputation or synthesis of potentially unavailable FLAIR sequences.
Quantification of white matter lesions
State-of-the-art brain segmentation using deep learning. The underlying deep-learning based versatile segmentation engine resulted in a top-ranked entry for the recent Medical Segmentation Decathlon Grand Challenge.
Predicting and distinguishing between Alzheimer’s and Alzheimer’s subtypes
Moving from Radius’ quantification of biomarkers and into focusing on our first product Apollo, 2019 opened new opportunities in terms of validation of our patent pending technology.
Predicting and distinguishing between Alzheimer’s and Alzheimer’s subtypes became a main focus area.
The value of hippocampal volume, shape, and texture for 11-year prediction of dementia: a population-based study
HC Achterberg, L Sørensen, FJ Wolters, WJ Niessen, MW Vernooij, …
Neurobiology of aging 81, 58-66.
Unsupervised Machine Learning On Baseline Brain Mri Identifies Mci Subgroup With A Faster Decline Over Two Years Compared To Classical Hippocampal Sparing Ad Subtype
L Sørensen, A Pai, M Nielsen, JB Leverenz, JA Pillai
Alzheimer’s & Dementia: The Journal of the Alzheimer’s Association 15 (7), P1420
Training recurrent neural networks robust to incomplete data: Application to Alzheimer’s disease progression modeling
MM Ghazi, M Nielsen, A Pai, MJ Cardoso, M Modat, S Ourselin, …
Medical image analysis 53, 39-46
How can we cope with less training data?
Working with ML training of specified pathology identifications can be a challenge when the availability of case studies can be scarce. To combat this issue methods where developed to cope with this issue.
Multi-domain adaptation in brain MRI through paired consistency and adversarial learning
M Orbes-Arteaga, T Varsavsky, CH Sudre, Z Eaton-Rosen, LJ Haddow, …
Knowledge distillation for semi-supervised domain adaptation
M Orbes-Arteainst, J Cardoso, L Sørensen, C Igel, S Ourselin, M Modat, …
OR 2.0 Context-Aware Operating Theaters and Machine Learning in Clinical …
PADDIT: Probabilistic Augmentation of Data using Diffeomorphic Image Transformation
M Orbes-Arteaga, L Sørensen, J Cardoso, M Modat, S Ourselin, …
Medical Imaging 2019: Image Processing 10949, 109490S
In 2020 Cerebriu will continue our previous endeavors and also concentrate on long term research taking us to other body parts than the brain. Ongoing activity is in breast cancer and arthritis in the knee.