Two different visual stimuli that cause axial eye shortening have no additive effect.

Previous studies identified two visual stimuli that can shorten the human eye by thickening the choroid after short-term visual stimulation, potentially inhibiting myopia: (1) watching digitally filtered movies where the red plane has full spatial resolution while green and blue are low-pass filtered according to the human longitudinal chromatic aberration (LCA) function (the "red in focus" filter), and (2) reading text with inverted contrast. This study aimed to determine whether combining these two stimuli would have an additive effect on axial length. Twenty-two emmetropic subjects were recruited to read text (standard and inverted contrast) for 30 min from a large screen, 2 m away, either unfiltered or filtered with the "red in focus" filter. Axial length was measured before and after each reading episode using low-coherence interferometry (Lenstar LS 900, Haag Streit). Reading text with conventional contrast polarity (dark letters on a bright background) resulted in no significant axial length change. Adding the "red in focus" filter did not alter the outcome. Consistent with previous findings, reading inverted contrast text made emmetropic eyes shorter. Surprisingly, when the text was combined with the "red in focus" filter, eyes became longer rather than shorter. A possible explanation for this contradictory result is that, for the text stimulus, the "red in focus" filter removes spatial information in the blue channel needed by the retina to use LCA analysis to thicken the choroid.

Automated deep learning segmentation of neuritic plaques and neurofibrillary tangles in Alzheimer disease brain sections using a proprietary software.

Neuropathological diagnosis of Alzheimer disease (AD) relies on semiquantitative analysis of phosphorylated tau-positive neurofibrillary tangles (NFTs) and neuritic plaques (NPs), without consideration of lesion heterogeneity in individual cases. We developed a deep learning workflow for automated annotation and segmentation of NPs and NFTs from AT8-immunostained whole slide images (WSIs) of AD brain sections. Fifteen WSIs of frontal cortex from 4 biobanks with varying tissue quality, staining intensity, and scanning formats were analyzed. We established an artificial intelligence (AI)-driven iterative procedure to improve the generation of expert-validated annotation datasets for NPs and NFTs thereby increasing annotation quality by >50%. This strategy yielded an expert-validated annotation database with 5013 NPs and 5143 NFTs. We next trained two U-Net convolutional neural networks for detection and segmentation of NPs or NFTs, achieving high accuracy and consistency (mean Dice similarity coefficient: NPs, 0.77; NFTs, 0.81). The workflow showed high generalization performance across different cases. This study serves as a proof-of-concept for the utilization of proprietary image analysis software (Visiopharm) in the automated deep learning segmentation of NPs and NFTs, demonstrating that AI can significantly improve the annotation quality of complex neuropathological features and enable the creation of highly precise models for identifying these markers in AD brain sections.