Search

Background: Alzheimer's disease (AD) is a heterogeneous, multifactorial neurodegenerative disorder characterized by three neurobiological factors beta‐amyloid, pathologic tau, and neurodegeneration. There are no effective treatments for AD at a late stage, urging for early detection and prevention. However, existing statistical inference approaches in neuroimaging studies of AD subtype identification do not take into account the pathological domain knowledge, which could lead to ill‐posed results that are sometimes inconsistent with the essential neurological principles. Purpose: Integrating systems biology modeling with machine learning, the study aims to assist clinical AD prognosis by providing a subpopulation classification in accordance with essential biological principles, neurological patterns, and cognitive symptoms. Methods: We propose a novel pathology steered stratification network (PSSN) that incorporates established domain knowledge in AD pathology through a reaction‐diffusion model, where we consider non‐linear interactions between major biomarkers and diffusion along the brain structural network. Trained on longitudinal multimodal neuroimaging data, the biological model predicts long‐term evolution trajectories that capture individual characteristic progression pattern, filling in the gaps between sparse imaging data available. A deep predictive neural network is then built to exploit spatiotemporal dynamics, link neurological examinations with clinical profiles, and generate subtype assignment probability on an individual basis. We further identify an evolutionary disease graph to quantify subtype transition probabilities through extensive simulations. Results: Our stratification achieves superior performance in both inter‐cluster heterogeneity and intra‐cluster homogeneity of various clinical scores. Applying our approach to enriched samples of aging populations, we identify six subtypes spanning AD spectrum, where each subtype exhibits a distinctive biomarker pattern that is consistent with its clinical outcome. Conclusions: The proposed PSSN (i) reduces neuroimage data to low‐dimensional feature vectors, (ii) combines AT[N]‐Net based on real pathological pathways, (iii) predicts long‐term biomarker trajectories, and (iv) stratifies subjects into fine‐grained subtypes with distinct neurological underpinnings. PSSN provides insights into pre‐symptomatic diagnosis and practical guidance on clinical treatments, which may be further generalized to other neurodegenerative diseases. [ABSTRACT FROM AUTHOR]

Purpose: Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) shows high sensitivity in detecting breast cancer. However, its performance could be affected by patient motion during the imaging. To overcome this problem, it is necessary to correct patient motion by deformable registration, before using the DCE-MRI to detect breast cancer. However, deformable registration of DCE-MR images is challenging due to the dramatic contrast change over time (especially between the precontrast and postcontrast images). Most existing methods typically register each postcontrast image onto the precontrast image independently, without considering the dynamic contrast change after agent uptake. This could lead to the inconsistency among the aligned postcontrast images in the precontrast image space, which will eventually result in worse performance in cancer detection. In this paper, the authors present a novel hierarchical registration framework to address this problem. Methods: First, the authors propose a hierarchical registration framework to deploy the groupwise registration for simultaneous registration of all postcontrast images onto their group-mean image and further aligning the group-mean image of postcontrast images onto the precontrast image space for final alignment of all precontrast and postcontrast images. In this way, the postcontrast images (with similar intensity patterns) can be jointly aligned onto the precontrast image for increasing their overall consistency after registration. Second, in order to improve the registration between the precontrast image and the group-mean image of the postcontrast images, the authors propose using the contrast-invariant attribute vectors to guide the robust feature matching during the registration. Results: Our proposed hierarchical registration framework has been comprehensively evaluated and compared with affine registration and widely used deformable registration methods in both pairwise and groupwise registration formulation. The experimental results on both real and simulated images show that our method can obtain not only more accurate but also more consistent registration results than any of all other registration algorithms.Conclusions: The authors have proposed a novel groupwise registration method to achieve accurate and consistent alignment for breast DCE-MR images. In the future, the authors will further evaluate our proposed method with more clinical datasets. [ABSTRACT FROM AUTHOR]