Deep learning models for neuroimaging have largely been developed for individual tasks, limiting knowledge transfer across applications. Here we introduce GenFAR, a modular deep learning framework that learns general, clinically informed features from brain MRIs. We trained this modular architecture on 49,246 individuals across 11 cohorts, using 17 diverse classification and regression tasks spanning cognition, clinical, diagnosis, demographics, and biomarkers. This yields aggregated, focused feature sets that capture rich, clinically- and biologically-relevant brain representations. We developed a sequential learning approach where tasks progressively build on previously learned representations. Through an analysis of 5,000 task sequences, we identified an optimal sequence length of six tasks and introduced a Donor Score metric to quantify each task's contribution to downstream performance. This analysis revealed five consistently strong donor tasks (Age, AD/MCI, MMSE, Hypertension, Hyperlipidemia) that formed the base of our sequential model. We demonstrated the utility of our learned representation, in various tasks beyond those included in the training set, to serve as the foundation for specialized secondary predictors. We further showed that using the learned feature representation can substantially increase the sample efficiency of secondary deep learning training tasks and models, as well as improve their accuracy.
V. Bashyam, G. Erus, Junhao Wen et al.· 0 citations
Importance: The long-term efficacy of amyloid targeting therapies hinge on their ability to slow downstream neuropathologic change, but little is known about the influence of amyloid clearance on tau pathology and neurodegeneration. Objective: To determine the post-mortem and in vivo association between post-treatment amyloid levels and downstream neuropathology in a patient with patchy areas showing minimal residual amyloid following aducanumab therapy. Design, Setting, and Participants: Clinicopathologic case report from a single academic memory center. A p.R47H TREM2 (a variant associated with higher Alzheimer disease risk) male carrier in his 50s with mild cognitive impairment who received aducanumab in the EMERGE/EMBARK trials and 14 age- or TREM2-matched untreated controls from the Penn Center for Neurodegenerative Disease Research. Exposures: 30 doses of aducanumab (cumulative dose 280mg/kg) over 4.5 years. Main Outcomes and Measures: Neuropathologic evaluation of amyloid, tau, and neuroinflammation; Amyloid PET and Tau PET standardized uptake value ratio, longitudinal change in cortical thickness. Results: Four years after receiving his final dose of aducanumab, the patient died and autopsy demonstrated variable levels of amyloid pathology, including regions with very low amyloid juxtaposed with regions showing typical high amyloid burden in deep cortical layers with only low amyloid burden in superficial layers. Regions showing low post-treatment amyloid were preferentially found in gyral crests and were associated with less tau pathology than untreated controls on autopsy and slower longitudinal atrophy on in vivo MRI (β = −0.50, [−0.62, −0.37], t = −7.96, p < .001). In contrast, regions with high amyloid burden were preferentially found in sulcal depths and had similar levels of tau pathology as seen in untreated controls on autopsy. Conclusion and Relevance: In this case report, areas of extensive amyloid clearance following amyloid targeting therapy were associated with less downstream neuropathological change and appear to preferentially occur in gyral crests. Future studies should evaluate the differential mechanisms involved in amyloid clearance from superficial and deep cortical layers and in gyri and sulci, as extensive amyloid clearance may be necessary to achieve downstream neuropathologic benefit following amyloid removal.
C. A. Brown, J. Robinson, Sandhitsu R. Das et al.· Journal of the American Medi...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
