CHIASM-Net: Artificial Intelligence-Based Direct Identification of Chiasmal Abnormalities in Albinism

Purpose

Albinism is a congenital disorder affecting pigmentation levels, structure, and function of the visual system. The identification of anatomical changes typical for people with albinism (PWA), such as optic chiasm malformations, could become an important component of diagnostics. Here, we tested an application of convolutional neural networks (CNNs) for this purpose.

Methods

We established and evaluated a CNN, referred to as CHIASM-Net, for the detection of chiasmal malformations from anatomic magnetic resonance (MR) images of the brain. CHIASM-Net, composed of encoding and classification modules, was developed using MR images of controls (n = 1708) and PWA (n = 32). Evaluation involved 8-fold cross validation involving accuracy, precision, recall, and F1-score metrics and was performed on a subset of controls and PWA samples excluded from the training. In addition to quantitative metrics, we used Explainable AI (XAI) methods that granted insights into factors driving the predictions of CHIASM-Net.

Results

The results for the scenario indicated an accuracy of 85 ± 14%, precision of 90 ± 14% and recall of 81 ± 18%. XAI methods revealed that the predictions of CHIASM-Net are driven by optic-chiasm white matter and by the optic tracts.

Conclusions

CHIASM-Net was demonstrated to use relevant regions of the optic chiasm for albinism detection from magnetic resonance imaging (MRI) brain anatomies. This indicates the strong potential of CNN-based approaches for visual pathway analysis and ultimately diagnostics.

Deep Learning-Based Detection of Malformed Optic Chiasms From MRI Images

Convolutional neural network (CNN) models are of great promise to aid the segmentation and analysis of brain structures. Here, we tested whether CNN trained to segment normal optic chiasms from the T1w magnetic resonance imaging (MRI) image can be also applied to abnormal chiasms, specifically with optic nerve misrouting as typical for human albinism. We performed supervised training of the CNN on the T1w images of control participants (n = 1049) from the Human Connectome Project (HCP) repository and automatically generated algorithm-based optic chiasm masks. The trained CNN was subsequently tested on data of persons with albinism (PWA; n = 9) and controls (n = 8) from the CHIASM repository. The quality of outcome segmentation was assessed via the comparison to manually defined optic chiasm masks using the Dice similarity coefficient (DSC). The results revealed contrasting quality of masks obtained for control (mean DSC ± SEM = 0.75 ± 0.03) and PWA data (0.43 ± 0.8, few-corrected p = 0.04). The fact that the CNN recognition of the optic chiasm fails for chiasm abnormalities in PWA underlines the fundamental differences in their spatial features. This finding provides proof of concept for a novel deep-learning-based diagnostics approach of chiasmal misrouting from T1w images, as well as further analyses on chiasmal misrouting and their impact on the structure and function of the visual system.

[Neuro-computational approaches for objective assessment of visual function]

BACKGROUND

Retinal pathologies affect the structure and function of post-retinal visual pathways. These post-retinal alterations bear the potential to obstruct the aim of innovative retinal treatment to restore visual function.

OBJECTIVE

Current developments in the field of neuroimaging and the associated neurocomputational approaches enable a detailed assessment of this interrelationship. As a consequence, they open up the possibility to anticipate the success of treatment.

METHODS

This review article demonstrates how innovations particularly in magnetic resonance imaging (MRI)-based anatomical, functional, and diffusion imaging can guide visual pathway assessments that are relevant for ophthalmological applications.

RESULTS

Specific examples of retinal and visual pathway pathologies in the context of a detailed analysis of the visual pathway are described.

CONCLUSION

A concept is introduced of how to translate the meaningful but technically and computationally challenging neuroimaging procedures into a clinical setting in order to effectively connect these procedures to innovative treatment approaches.

Triple visual hemifield maps in a case of optic chiasm hypoplasia

In humans, each hemisphere comprises an overlay of two visuotopic maps of the contralateral visual field, one from each eye. Is the capacity of the visual cortex limited to these two maps or are plastic mechanisms available to host more maps? We determined the cortical organization of the visual field maps in a rare individual with chiasma hypoplasia, where visual cortex plasticity is challenged to accommodate three hemifield maps. Using high-resolution fMRI at 7T and diffusion-weighted MRI at 3T, we found three hemiretinal inputs, instead of the normal two, to converge onto the left hemisphere. fMRI-based population receptive field mapping of the left V1-V3 at 3T revealed three superimposed hemifield representations in the left visual cortex, i.e. two representations of opposing visual hemifields from the left eye and one right hemifield representation from the right eye. We conclude that developmental plasticity including the re-wiring of local intra- and cortico-cortical connections is pivotal to support the coexistence and functioning of three hemifield maps within one hemisphere.

Quantifying nerve decussation abnormalities in the optic chiasm

•

Diffusion MRI is capable of detecting structural abnormalities of the optic chiasm.

•

Quantification of crossing strength in optic chiasm is of promise for albinism diagnostics.

•

Optic chiasm is a powerful test model for neuroimaging methods resolving crossing fibers.

Objective

The human optic chiasm comprises partially crossing optic nerve fibers. Here we used diffusion MRI (dMRI) for the in-vivo identification of the abnormally high proportion of crossing fibers found in the optic chiasm of people with albinism.

Methods

In 9 individuals with albinism and 8 controls high-resolution 3T dMRI data was acquired and analyzed with a set of methods for signal modeling [Diffusion Tensor (DT) and Constrained Spherical Deconvolution (CSD)], tractography, and streamline filtering (LiFE, COMMIT, and SIFT2). The number of crossing and non-crossing streamlines and their weights after filtering entered ROC-analyses to compare the discriminative power of the methods based on the area under the curve (AUC). The dMRI results were cross-validated with fMRI estimates of misrouting in a subset of 6 albinotic individuals.

Results

We detected significant group differences in chiasmal crossing for both unfiltered DT (p = 0.014) and CSD tractograms (p = 0.0009) also reflected by AUC measures (for DT and CSD: 0.61 and 0.75, respectively), underlining the discriminative power of the approach. Estimates of crossing strengths obtained with dMRI and fMRI were significantly correlated for CSD (R² = 0.83, p = 0.012). The results show that streamline filtering methods in combination with probabilistic tracking, both optimized for the data at hand, can improve the detection of crossing in the human optic chiasm.

Conclusions

Especially CSD-based tractography provides an efficient approach to detect structural abnormalities in the optic chiasm. The most realistic results were obtained with filtering methods with parameters optimized for the data at hand.

Significance

Our findings demonstrate a novel anatomy-driven approach for the individualized diagnostics of optic chiasm abnormalities.