Noah Charles Benson

Peripheral vision is crucial for daily activities and quality of life, yet traditional measures of visual function like visual acuity primarily assess central vision. Visual field tests can evaluate peripheral vision but require extended focus and precise fixation, often challenging for patients with severe sight loss. Functional MRI (fMRI) with population receptive field (pRF) mapping offers a non-invasive way to map scotomas but relies on single contrast levels and accurate fixation. We developed an fMRI-based approach to measure contrast sensitivity across the visual field without requiring precise fixation. Combining large-field stimulation with varying spatial frequencies and contrast levels, and either pRF mapping or a structure-based retinotopic atlas, we modeled contrast sensitivity in the primary visual cortex (V1) over a large (40 deg) expanse of the visual field. In seven normally sighted participants, we characterized differences in V1 sensitivity across eccentricities and visual quadrants, finding reliable and reproducible patterns at individual and session levels. To test robustness to fixation variability, we investigated how varying levels of eye movement affected V1 sensitivity patterns in two participants: cortical sensitivity patterns were largely preserved despite eye movements, particularly at low spatial frequencies. This suggests our approach can accommodate several degrees of fixation instability, making it suitable for populations with unstable or biased fixation for whom visual field maps are harder to acquire behaviorally (e.g., patients with dense central scotoma or strabismus). Additionally, our method effectively visualized simulated and disease-linked sensitivity loss at the cortical level. Crucially, these results could be largely recovered using the structure-based retinotopic atlas, eliminating the need for pRF mapping and precise fixation, although with reduced sensitivity. This approach, integrating large-field stimulation with a retinotopic atlas, offers a promising tool for monitoring vision loss and recovery across a range of visual impairments, addressing a significant challenge in current clinical assessments.

Isothermal nucleic acid amplification tests enable rapid and decentralized molecular diagnostics but often lack robust quantitative readouts compared to quantitative PCR. Here, we present a semi-quantitative nucleic acid measurement approach using machine learning to extract spatiotemporal features from real-time fluorescence imaging of rapid isothermal amplification reactions in microfluidic chips. A convolutional neural network was trained on multiple images sampled throughout a chip-based recombinase polymerase amplification reaction to classify samples into clinically relevant or logarithmically spaced concentration ranges spanning five orders of magnitude. The clinical classification model achieved 94.6% accuracy, and the logarithmic model achieved 92.7% accuracy, with most errors occurring between adjacent concentration categories. By learning spatiotemporal patterns of fluorescence development rather than relying on explicit feature extraction, the model remained accurate at both high and low nucleic acid concentration regimes where other quantitative isothermal molecular tests struggle. This approach enables automated interpretation of amplification reactions and extends the usable dynamic range of the assay. These results demonstrate that integrating machine learning with image-based amplification methods can support rapid semi-quantitative molecular testing and may facilitate broader deployment of nucleic acid diagnostics outside centralized laboratory settings.

Peripheral vision is crucial for daily activities and quality of life, yet traditional measures of visual function like visual acuity primarily assess central vision. Visual field tests can evaluate peripheral vision but require extended focus combined with precise fixation, often very challenging for patients with severe sight loss. Functional MRI (fMRI) with population receptive field (pRF) mapping offers a non-invasive way to map scotomas but is limited by its reliance on single contrast levels and the necessity of accurate fixation. We developed an fMRI-based approach to measure contrast sensitivity across the visual field without the need for precise fixation. By combining large-field stimulation with varying spatial frequencies and contrast levels with either pRF mapping or a retinotopic atlas based on anatomical landmarks, we modeled contrast sensitivity in the primary visual cortex (V1) over a large (40 deg) expanse of the visual field. In seven normal-sighted participants, we characterized differences in V1 cortical sensitivity across eccentricities and visual quadrants, finding reliable and reproducible patterns of sensitivity differences at individual and session levels. To assess the method’s tolerance to fixation variability, we further investigated how different levels of eye movement affect cortical sensitivity patterns in two participants. We found that cortical sensitivity patterns were largely preserved across eye movement, particularly at low spatial frequencies. This suggests that our approach can accommodate several degrees of fixation instability, making it suitable for populations with unstable or biased fixation for whom visual field maps are harder to acquire behaviorally (e.g., patients with dense central scotoma or strabismus). Additionally, our method effectively visualized cases of simulated and disease-linked sensitivity loss at the cortical level. Crucially, we demonstrated that these results could be largely recovered using a structure-based retinotopic atlas, eliminating the need for pRF mapping and precise fixation - although such an approach reduced sensitivity. This approach, integrating large-field stimulation with a retinotopic atlas, offers a promising tool for monitoring vision loss and recovery in patients with various visual impairments, addressing a significant challenge in current clinical assessments.

Topographic organization, whereby neighboring cortical locations encode neighboring features in sensory or cognitive space, is a fundamental principle of brain function. Existing approaches for obtaining individual-specific topographic maps either require resource-intensive functional neuroimaging or, when relying on population atlases, lack precision for individual-level inference. Here, we introduce deepRetinotopy toolbox, a deep learning-based application for predicting the functional topographic organization of human visual cortex from cortical anatomy alone. DeepRetinotopy toolbox produces accurate retinotopic maps across diverse experimental conditions, imaging sites, and scanner types. We demonstrate how predicted maps can be utilized to automatically generate individual-specific visual area boundaries, overcoming common biases in manual annotations. Finally, we applied our method to 11,060 anatomical scans, which allowed us to quantify age-related changes in the functional organization of visual cortex predictable from anatomy alone, underscoring the method’s broad utility for scalable, anatomy-based functional brain mapping.

Topographic organization, whereby neighboring cortical locations encode neighboring features in sensory or cognitive space, is a fundamental principle of brain function. Existing approaches for obtaining individual-specific topographic maps either require resource-intensive functional neuroimaging or, when relying on population atlases, lack precision for individual-level inference. Here, we introduce deepRetinotopy toolbox , a deep learning-based application for predicting the functional topographic organization of human visual cortex from cortical anatomy alone. DeepRetinotopy toolbox produces accurate retinotopic maps across diverse experimental conditions, imaging sites, and scanner types. We demonstrate how predicted maps can be utilized to automatically generate individual-specific visual area boundaries, overcoming common biases in manual annotations. Finally, we applied our method to 11,060 anatomical scans, which allowed us to quantify age-related changes in the functional organization of visual cortex predictable from anatomy alone, underscoring the method’s broad utility for scalable, anatomy-based functional brain mapping.

The visual cortex encodes the visual field through numerous bilaterally paired, topologically organized two-dimensional maps along the cortical surface. Although these representations exhibit a largely consistent organization across individuals, substantial interindividual variability exists in both the structure and functional organization of the visual cortex. To better characterize this variability, researchers have increasingly turned to computational approaches, alongside empirical methodologies and encoding models, to predict individual-level retinotopic organization. As these advances continue to shape the study of retinotopic organization, it is crucial to revisit the methodologies used to generate and model these maps. In this review, we examine empirical and theoretical work aimed at reconstructing and modeling visual field maps in the human visual cortex. Specifically, we discuss how empirical retinotopic mapping has facilitated the development of theoretical and computational models of retinotopy and, in turn, how these models have enhanced our understanding of the retinotopic organization of the human visual cortex. Finally, we outline a non-exhaustive set of future directions for leveraging models of retinotopy to further investigate structure-function relationships, interindividual variability, and more.

A major problem in human visual neuroscience research is the localization of visual areas on the cortical surface. Currently available methods are capable of making detailed predictions about many areas, but human raters do not agree as well with these methods as they do with each other. Although highly accurate, human raters require substantial time and expertise that many researchers do not have. Additionally, human raters require functional data for drawing visual area boundaries that requires additional scan time, budget, and expertise to collect. Here, we train convolutional neural network (CNN) models to predict the boundaries and iso-eccentric regions of V1, V2, and V3 in both the Human Connectome Project dataset and the NYU Retinotopy dataset. CNNs trained to use the same functional data available to human raters predicted these boundaries with an accuracy similar to humans, while CNNs trained to use only anatomical data had a lower accuracy that was nonetheless higher than that of any currently available method. A comparison of the model accuracies when predicting eccentricity-based boundaries and polar angle-based boundaries suggests that eccentricity is substantially less closely tied to anatomical structure than polar angle and that the cortical magnification function, at least in terms of eccentricity, varies substantially between subjects. We further find that the fraction of V1, V2, and V3 that can be accurately parcellated into function regions using gray matter structural data alone is ∼75% (∼80% of the inter-rater reliability of human experts), implying a much tighter coupling between structure and function in these areas than previously estimated. We conclude that machine learning techniques such as CNNs provide a powerful tool for mapping the brain with human accuracy and predict that such tools will become integral to neuroscience research going forward.

Visual recognition is limited by both object size (acuity) and spacing. The spacing limit, called “crowding”, is the failure to recognize an object in the presence of other objects. Here, we take advantage of individual differences in crowding to investigate its biological basis. Crowding distance, the minimum object spacing needed for recognition, varies 2-fold among healthy adults. We test the conjecture that this variation in psychophysical crowding distance is due to variation in cortical map size. To test this, we make paired measurements of brain and behavior in 49 observers. We use psychophysics to measure crowding distance and calculate λ, the number of letters that fit into each observer’s visual field without crowding. In the same observers, we use functional magnetic resonance imaging (fMRI) to measure the surface area A of retinotopic maps V1, V2, V3, and V4. Across observers, λ is proportional to the surface area of V4 but is uncorrelated with the surface area of V1 to V3. The proportional relationship of λ to area of V4 indicates conservation of cortical crowding distance across individuals: letters can be recognized if they are spaced by at least 1.4 mm on the V4 map, irrespective of map size and psychophysical crowding distance. We conclude that the size of V4 predicts the spacing limit of visual perception.

Perceptual brain regions represent the world using 2D maps. In the visual cortex in particular, the 2D cortical surface represents the 2D visual field topologically. That is, adjacent neurons in the visual brain represent adjacent locations on the retina, resembling a miniature map of the outside world. Here, we review empirical and theoretical work to access and model these miniature maps in human visual cortex. We aim to present a balanced view of the foundational work of the past and the current state-of-the-art capabilities. Specifically, we discuss how empirical retinotopic mapping enabled the development of theoretical and computational models of retinotopy and, subsequently, how these models have been used to better understand the retinotopic organization of the human visual cortex. Finally, we discuss a non-exhaustive list of potential future directions to further our understanding of these visual field maps, including their structure-functional relationships, individual variability, and more.

Peripheral vision is crucial for daily activities and quality of life, yet traditional measures of visual function like visual acuity primarily assess central vision. Visual field tests can evaluate peripheral vision but require extended focus combined with precise fixation, often very challenging for patients with severe sight loss. Functional MRI (fMRI) with population receptive field (pRF) mapping offers a non-invasive way to map scotomas but is limited by its reliance on single contrast levels and the necessity of accurate fixation. We developed an fMRI-based approach to measure contrast sensitivity across the visual field without the need for precise fixation. By combining large-field stimulation with varying spatial frequencies and contrast levels with either pRF mapping or a retinotopic atlas based on anatomical landmarks, we modeled contrast sensitivity in the primary visual cortex (V1) over a large (40 deg) expanse of the visual field. In seven normal-sighted participants, we characterized differences in V1 cortical sensitivity across eccentricities and visual quadrants, finding reliable and reproducible patterns of sensitivity differences at individual and session levels. To assess the method’s tolerance to fixation variability, we further investigated how different levels of eye movement affect cortical sensitivity patterns in two participants. We found that cortical sensitivity patterns were largely preserved across eye movement, particularly at low spatial frequencies. This suggests that our approach can accommodate several degrees of fixation instability, making it suitable for populations with unstable or biased fixation for whom visual field maps are harder to acquire behaviorally (e.g., patients with dense central scotoma or strabismus). Additionally, our method effectively visualized cases of simulated and disease-linked sensitivity loss at the cortical level. Crucially, we demonstrated that these results could be largely recovered using a structure-based retinotopic atlas, eliminating the need for pRF mapping and precise fixation - although such an approach reduced sensitivity. This approach, integrating large-field stimulation with a retinotopic atlas, offers a promising tool for monitoring vision loss and recovery in patients with various visual impairments, addressing a significant challenge in current clinical assessments.