Search

Scene recognition is an essential component of both machine and biological vision. Recent advances in computer vision using deep convolutional neural networks (CNNs) have demonstrated impressive sophistication in scene recognition, through training on large datasets of labeled scene images (Zhou et al. 2018, 2014). One criticism of CNN-based approaches is that performance may not generalize well beyond the training image set (Torralba and Efros 2011), and may be hampered by minor image modifications, which in some cases are barely perceptible to the human eye (Goodfellow et al. 2015; Szegedy et al. 2013). While these “adversarial examples” may be unlikely in natural contexts, during many real-world visual tasks scene information can be degraded or limited due to defocus blur, camera motion, sensor noise, or occluding objects. Here, we quantify the impact of several image degradations (some common, and some more exotic) on indoor/outdoor scene classification using CNNs. For comparison, we use human observers as a benchmark, and also evaluate performance against classifiers using limited, manually selected descriptors. While the CNNs outperformed the other classifiers and rivaled human accuracy for intact images, our results show that their classification accuracy is more affected by image degradations than human observers. On a practical level, however, accuracy of the CNNs remained well above chance for a wide range of image manipulations that disrupted both local and global image statistics. We also examine the level of image-by-image agreement with human observers, and find that the CNNs' agreement with observers varied as a function of the nature of image manipulation. In many cases, this agreement was not substantially different from the level one would expect to observe for two independent classifiers. Together, these results suggest that CNN-based scene classification techniques are relatively robust to several image degradations. However, the pattern of classifications obtained for ambiguous images does not appear to closely reflect the strategies employed by human observers.

Introduction: Biomarkers for the prediction of cognitive decline in patients with amnestic mild cognitive impairment (MCI) and amnestic mild dementia are needed for both clinical practice and clinical trials. Methods: We evaluated the ability of tau‐PET (positron emission tomography), cortical atrophy on magnetic resonance imaging (MRI), baseline cognition, apolipoprotein E gene (APOE) status, plasma and cerebrospinal fluid (CSF) levels of phosphorylated tau‐217, neurofilament light (NfL), and amyloid beta (Aβ)42/40 ratio (individually and in combination) to predict cognitive decline over 2 years in BioFINDER‐2 and Alzheimer's Disease Neuroimaging Initiative (ADNI). Results: Baseline tau‐PET and a composite baseline cognitive score were the strongest independent predictors of cognitive decline. Cortical thickness and NfL provided some additional information. Using a predictive algorithm to enrich patient selection in a theoretical clinical trial led to a significantly lower required sample size. Discussion: Models including baseline tau‐PET and cognition consistently provided the best prediction of change in cognitive function over 2 years in patients with amnestic MCI or mild dementia. [ABSTRACT FROM AUTHOR]

Introduction: The effect of random error on the performance of blood‐based biomarkers for Alzheimer's disease (AD) must be determined before clinical implementation. Methods: We measured test‐retest variability of plasma amyloid beta (Aβ)42/Aβ40, neurofilament light (NfL), glial fibrillary acidic protein (GFAP), and phosphorylated tau (p‐tau)217 and simulated effects of this variability on biomarker performance when predicting either cerebrospinal fluid (CSF) Aβ status or conversion to AD dementia in 399 non‐demented participants with cognitive symptoms. Results: Clinical performance was highest when combining all biomarkers. Among single‐biomarkers, p‐tau217 performed best. Test‐retest variability ranged from 4.1% (Aβ42/Aβ40) to 25% (GFAP). This variability reduced the performance of the biomarkers (≈ΔAUC [area under the curve] −1% to −4%) with the least effects on models with p‐tau217. The percent of individuals with unstable predicted outcomes was lowest for the multi‐biomarker combination (14%). Discussion: Clinical prediction models combining plasma biomarkers—particularly p‐tau217—exhibit high performance and are less effected by random error. Individuals with unstable predicted outcomes ("gray zone") should be recommended for further tests. [ABSTRACT FROM AUTHOR]

Objective: To investigate the statistical power of plasma, imaging, and cognition biomarkers as Alzheimer's disease (AD) clinical trial outcome measures. Methods: Plasma neurofilament light, structural magnetic resonance imaging, and cognition were measured longitudinally in the Alzheimer's Disease Neuroimaging Initiative (ADNI) in control (amyloid PET or CSF Aβ42 negative [Aβ‐] with Clinical Dementia Rating scale [CDR] = 0; n = 330), preclinical AD (Aβ + with CDR = 0; n = 218) and mild AD (Aβ + with CDR = 0.5‐1; n = 697) individuals. A statistical power analysis was performed across biomarkers and groups based on longitudinal mixed effects modeling and using several different clinical trial designs. Results: For a 30‐month trial of preclinical AD, both the temporal composite and hippocampal volumes were superior to plasma neurofilament light and cognition. For an 18‐month trial of mild AD, hippocampal volume was superior to all other biomarkers. Plasma neurofilament light became more effective with increased trial duration or sampling frequency. Imaging biomarkers were characterized by high slope and low within‐subject variability, while plasma neurofilament light and cognition were characterized by higher within‐subject variability. Interpretation: MRI measures had properties that made them preferable to cognition and pNFL as outcome measures in clinical trials of early AD, regardless of cognitive status. However, pNfL and cognition can still be effective depending on inclusion criteria, sampling frequency, and response to therapy. Future trials will help to understand how sensitive pNfL and MRI are to detect downstream effects on neurodegeneration of drugs targeting amyloid and tau pathology in AD. [ABSTRACT FROM AUTHOR]

Introduction: Biomarkers for the prediction of cognitive decline in patients with amnestic mild cognitive impairment (MCI) and amnestic mild dementia are needed for both clinical practice and clinical trials., Methods: We evaluated the ability of tau-PET (positron emission tomography), cortical atrophy on magnetic resonance imaging (MRI), baseline cognition, apolipoprotein E gene (APOE) status, plasma and cerebrospinal fluid (CSF) levels of phosphorylated tau-217, neurofilament light (NfL), and amyloid beta (Aβ) 42/40 ratio (individually and in combination) to predict cognitive decline over 2 years in BioFINDER-2 and Alzheimer's Disease Neuroimaging Initiative (ADNI)., Results: Baseline tau-PET and a composite baseline cognitive score were the strongest independent predictors of cognitive decline. Cortical thickness and NfL provided some additional information. Using a predictive algorithm to enrich patient selection in a theoretical clinical trial led to a significantly lower required sample size., Discussion: Models including baseline tau-PET and cognition consistently provided the best prediction of change in cognitive function over 2 years in patients with amnestic MCI or mild dementia., (© 2022 The Authors. Alzheimer's & Dementia published by Wiley Periodicals LLC on behalf of Alzheimer's Association.)

Introduction: The effect of random error on the performance of blood-based biomarkers for Alzheimer's disease (AD) must be determined before clinical implementation., Methods: We measured test-retest variability of plasma amyloid beta (Aβ)42/Aβ40, neurofilament light (NfL), glial fibrillary acidic protein (GFAP), and phosphorylated tau (p-tau)217 and simulated effects of this variability on biomarker performance when predicting either cerebrospinal fluid (CSF) Aβ status or conversion to AD dementia in 399 non-demented participants with cognitive symptoms., Results: Clinical performance was highest when combining all biomarkers. Among single-biomarkers, p-tau217 performed best. Test-retest variability ranged from 4.1% (Aβ42/Aβ40) to 25% (GFAP). This variability reduced the performance of the biomarkers (≈ΔAUC [area under the curve] -1% to -4%) with the least effects on models with p-tau217. The percent of individuals with unstable predicted outcomes was lowest for the multi-biomarker combination (14%)., Discussion: Clinical prediction models combining plasma biomarkers-particularly p-tau217-exhibit high performance and are less effected by random error. Individuals with unstable predicted outcomes ("gray zone") should be recommended for further tests., (© 2022 The Authors. Alzheimer's & Dementia published by Wiley Periodicals LLC on behalf of Alzheimer's Association.)