Application of deep learning to understand resilience to Alzheimer's disease pathology

Abstract

People who have Alzheimer's disease neuropathologic change (ADNC) typically associated with dementia but not the associated cognitive decline can be considered to be "resilient" to the effects of ADNC. We have previously reported lower neocortical levels of hyperphosphorylated tau (pTau) and less limbic-predominant age-related TDP-43 encephalopathy neuropathologic change (LATE-NC) in the resilient participants compared to those with dementia and similar ADNC as determined by current NIA-AA recommendations using traditional semi-quantitative assessments of amyloid β and pathological tau burden. To better understand differences between AD-dementia and resilient participants, we developed and applied a deep learning approach to ana- CON F L IC T OF I N T E R E STDr. A. Lee reports support from the US Food and Drug Administration, grants from Santen, Carl Zeiss Meditec, and Novartis, personal fees from Genentech, Topcon, and Verana Health, outside of the submitted work; this article does not reflect the opinions of the Food and Drug Administration. Dr. Larson receives royalty funds from UpToDate. No conflicting relationship exists for the rest of the authors. AU T HOR CON T R I BU T ION SCecilia S. Lee, and Aaron Y. Lee had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Retinal Biomarkers for Alzheimer Disease: The Facts and the Future

“…Potential applications of AI include providing new insights based on known features, or discerning novel features and biomarkers that are not currently known. As an example of the former, deep learning algorithms applied to histologic sections from brain biopsies of AD patients were able to successfully quantify levels of phosphorylated tau and neurofibrillary tangles, and found that neurofibrillary tangle burden was strongly associated with cognitive impairment in AD patients 69 . As an example of the latter, parallel work on identifying early biomarkers for age- related macular degeneration described a framework that was agnostic to prespecified parameters, and successfully discovered new structural biomarkers on OCT that correlated with function 70 …”

Role of Deep Learning in Predicting Aging-Related Diseases: A Scoping Review

“…We envision the review of joint use of DL over big data obtained from medical images, electronic medical records, research databases, personalized genomics and wearable sensor data for continuous monitoring of aging patients' health and risks of aforementioned diseases. DL has been successfully applied to imaging data to determine features that are associated with the detection of clinical diseases and also to more accurately classify disease progression in tissue-based studies [13]. DL has also been utilized over blood biochemistry parameters and cell count linked to chronological age with other factors such as sex and lifestyle [14].…”

The concept of resilience to Alzheimer’s Disease: current definitions and cellular and molecular mechanisms

“… Kobayashi et al, 2018 [ 83 ] Reactive astrocytes Entorhinal cortex 10 GLT-1 reactive astrocytes are associated with a preserved cognition in the presence of AD pathology CERAD non-frequent, Braak III-V Cognition based on CDR, (control and resilient = 0 – 0,5, AD = 2–3). Pathology well matched between AD and resilient cases Latimer et al, 2019 [ 84 ] ptau burden and LATE-NC Many cerebral regions and dentate gyrus 7 Resilient donors exhibited a lower cortical ptau burden, plaque burden and less LATE-NC CERAD, frequent, Braak IV Resilience is matched to demented patients with a similar ABC score, but demented patients show a significant increase in LATE-NC and macroinfarcts, cognition tested every two years with CASI Lee et al, 2019 [ 85 ] Network activity 116 different regions Network efficiency related to resilience and the right middle-temporal pole might center for neural substrates Based on Tau and Aβ PET Network efficiency related to resilience determined by residual approach of resilience (estimating difference between actual and estimated performance of cognition based on pathology, atrophy) Lee et al, 2021 [ 86 ] ptau burden by deep learning technique Medial frontal cortex and amygdala 7 Resilient donors exhibited a lower cortical ptau burden (despite ptau in neurites) and reduced LATE-NC CERAD frequent, Braak VI CERAD and Braak matched between AD and resilient donors, cognition based on CASI within 2 years of death Maarouf et al, 2011 [ 87 ] Biochemical assessment Frontal cortex 8 Resilient individuals lack Aβ-related biochemical changes and the authors suggested that not only Aβ but additionally ptau and microvascular dysfunction play a role in cognitive deterioration CERAD, moderate – frequent, Braak III-V Did not included healthy controls as reference MMSE ≥ 27 for resilient donors and ≤ 19 for the...…”

“…Strikingly, reduced amounts of NPs have also been observed in the superior temporal sulcus in resilient individuals compared to AD patients [ 91 ]. Similarly, in resilient donors with similar Braak stages compared to AD patients, reduced amounts of ptau has also been shown in the middle superior gyrus [ 84 ] and in NFTs in the temporal gyrus [ 86 ]. Also, the progression of ptau could be halted, as in nondemented centenarians very few resilient donors reached Braak stage 5 or a CERAD score for NPs higher than 2 [ 24 ].…”