Skip to main content

Accelerating Neuroscience Research by Unifying Knowledge Representation and Analysis Through a Domain Specific Language

Periodic Reporting for period 2 - NeuroLang (Accelerating Neuroscience Research by Unifying Knowledge Representation and Analysis Through a Domain Specific Language)

Reporting period: 2019-09-01 to 2021-02-28

Neuroscience is at an inflection point. The 150-year old cortical specialization paradigm, in which cortical brain areas have a distinct set of functions, is experiencing an unprecedented momentum with over 1000 articles being published every year. However, this paradigm is reaching its limits. Recent studies show that current approaches to atlas brain areas, like relative location, cellular population type, or connectivity, are not enough on their own to characterize a cortical area and its function unequivocally. This hinders the reproducibility and advancement of neuroscience.

Neuroscience is thus in dire need of a universal standard to specify neuroanatomy and function: a novel formal language allowing neuroscientists to simultaneously specify tissue characteristics, relative location, known function and connectional topology for the unequivocal identification of a given brain region.
The vision of NeuroLang is that a unified formal language for neuroanatomy will boost our understanding of the brain. By defining brain regions, networks, and cognitive tasks through a set of formal criteria, researchers will be able to synthesize and integrate data within and across diverse studies. NeuroLang will accelerate the development of neuroscience by providing a way to evaluate anatomical specificity, test current theories, and develop new hypotheses.
First, we proved that a large section of NeuroLang neuroanatomical DSL can inscribed in terms of a decidable extended segment of the Datalog+/- formalism. Our extension includes aggregation, negation, and interval algebras for spatial and temporal neuroimaging signal representation]. Our DSL can be found in http://neurolang.github.io

Second, we have used the logic programming paradigm to unify and formalize human sulcal neuroanatomy. Specifically, we have unified sulcal neuroanatomy across three main neuroanatomical schools by formalizing sulcal descriptions into our logical DSL. We harnessed these descriptions to automatically identify sulci in 500 human subjects analyzing spatial and topographical variability in the human brain. This led us to produce a novel taxonomy of human brain sulci based on their cross-subject stability. The methods and preliminary results have been published as abstracts [2, 6, 7] and a journal publication is in preparation. This development enabled our participation in a collaborative effort to produce a multi-scale atlas linking neuroanatomy and function [2]. Furthermore, we compared DSL-based automatic identification of white matter structures versus machine-learning based ones [1] showing that formalizing human neuroanatomical knowledge can provide better results for white matter tract identification when spatial agreement variability across humans. The integration of the white matter DSL into NeuroLang is on-going.

Third, based on the Datalog+/- formalism, we enabled our language to query information on graph representations brain structure and function, commonly used in systems neuroscience. Through this, NeuroLang, as proposed in the description of the action, seamlessly integrates heterogeneous datasets in open and close world formalism. Such datasets include tabular databases, neuroscience data such as structural and functional magnetic resonance images and connectivity graphs, and ontologies. Specifically, this enabled prototypical studies such as the integration of the NeuroSynth meta-analytic database, a tabular closed-world database; with the novel Jülich brain atlas, i.e. spatial information; functional and structural connectivity information on individual brains, information in the topology of graph; and ontologies such as the US National Library of Medicine’s NeuroFMA and Stanford’s cognitive atlas CogAt, which represent neuroanatomical and cognitive knowledge in terms of a knowledge graph in an open world formalism (article in preparation).

Fourth, we have developed an efficient extension of the Datalog+/- to probabilistic programming based on the CP-logic formalism, which allows us to model discrete random variables. Our current implementation of the probabilistic analysis is based on, when possible, safe query plans which guarantee a polynomial time/space computation, and alternatively, knowledge compilation [7]. A contribution in this aspect is showing that a large family of neuroscience-centered queries has a safe plan, and hence a polynomial algorithm provably exists. We have tested this implementation by generalizing meta-analytic analyses of neuroimaging data in terms of NeuroLang [7].

Fifth, our continuing collaboration with top-level neuroscience groups has enabled our team to participate in cognitive and neuroanatomical research [5, 8, 12] providing the grounds for adapting and developing NeuroLang in a close-knit relationship with the neuroscience community.

1.Sydnor V. J. et al. A comparison of three fiber tract delineation methods and their impact on white matter analysis. NeuroImage 178, 318–331 (2018).
2.Machlouzarides-Shalit A., Iovene, V., Makris, N. & Wassermann, D. A novel sulcal hierarchy based on manually labelled sulci. in (2019).
3.Gallardo G., Wassermann, D. & Anwander, A. Bridging the Gap: From Neuroanatomical Knowledge to Tractography of Brain Pathways. 2020 doi:10.1101/2020.08.01.232116.
4.Dadi K. et al. Fine-grain atlases of functional modes for fMRI analysis. NeuroImage 221, 117126 (2020).
5.Menon V. et al. Microstructural organization of human insula is linked to its macrofunctional circuitry and predicts cognitive control. eLife 9, e53470 (2020).
6.Machlouzarides-Shalit A. et al. NeuroLang: Representing neuroanatomy with sulcus-specific queries. in (2020).
7.Iovene V. & Wassermann, D. Probabilistic programming in neurolang: Bridging the gap between cognitive science and statistical modeling. in Organization for human brain mapping (2020).
8.Menon V. et al. Quantitative modeling links in vivo microstructural and macrofunctional organization of human and macaque insular cortex, and predicts cognitive control abilities. 2019 doi:10.1101/662601.
9.Wassermann D. et al. Sensing von economo neurons in the insula with multi-shell diffusion MRI. in (2018).
10.Gallardo G. et al. Solving the Cross-Subject Parcel Matching Problem Using Optimal Transport. in Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 (eds. Frangi, A. F., Schnabel, J. A., Davatzikos, C., Alberola-López, C. & Fichtinger, G.) vol. 11070 836–843 (Springer International Publishing, 2018).
11.Dockès, J. et al. Text to Brain: Predicting the Spatial Distribution of Neuroimaging Observations from Text Reports. in Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 (eds. Frangi, A. F., Schnabel, J. A., Davatzikos, C., Alberola-López, C. & Fichtinger, G.) vol. 11072 584–592 (Springer International Publishing, 2018).
12.Chen L. et al. The visual word form area (VWFA) is part of both language and attention circuitry. Nat Commun 10, 5601 (2019).
In all, we have shown that probabilistic logic programming (PLP) is a fit tool to formalise neuroanatomy, which is available at https://neurolang.github.io . Before the end of the project we expect to apply our PLP paradigm to the formalisation of neuroimaging-centered problems and, in doing this, to boost reproducibility and research in neuroimaging.
Example of Neurolang Applications for White Matter Anatomy