CNS*2023 Leipzig has ended
32nd Annual Computational Neuroscience Meeting
CNS*2023, Leipzig, Germany
Monday, July 17 • 16:40 - 18:40
P127: Proposal of an function-oriented SCID method for reverse engineering a wide range of brain regions

Log in to save this to your schedule, view media, leave feedback and see who's attending!

Brain Reference Architecture (BRA) data is a standardized representation of the computational processes and components involved in human brain functions. It can be used to guide the development of brain-inspired software12 and to understand neurological disorders1. BRA data consists of two parts: Brain Information Flow (BIF) diagram, which describes the mesoscopic-level anatomical structure of the brain, and Hypothetical Component Diagrams (HCDs), which aligns with BIF and specifies the functional roles and interactions of each component. The method for designing BRA data is called Structure-constrained Interface Decomposition (SCID). It involves three steps:
  • Step 1 BIF construction: Investigating brain anatomy and neural circuits in the Region of Interest (ROI) to define BIF.
  • Step 2 Alignment of ROI and TLF: Identifying the Top-Level Function (TLF).
  • Step 3 HCD design: From the HCDs, candidate component diagrams describing computational functions and interfaces are enumerated, and inappropriate ones are eliminated based on neuroscientific evidence and logical consistency. The remaining ones are HCDs.

As described above, the SCID method designs the computational function primarily from anatomical structures, which are relatively well-known in neuroscience. Given these technical circumstances, if BRA data are designed with a focus on local circuits, they can be applied in wide brain regions.

In the design of brain-inspired software, there is also a need to represent global computational functions such as survival and imaginative functions in BRA. However, it isn’t easy to create HCDs for broad regions using the SCID method. Defining BIFs and designing HCDs with an overall ROI is complex and difficult.

In this study, we propose a new methodology called the "Function-Oriented SCID method”. This methodology defines the computational function of a broad brain region as a TLF and creates an HCD in the following steps.
  • Step 1 TLF setting: The target function in a broad brain region (e.g., the whole brain) is specified as a TLF.
  • Step 2 Functional decomposition: The TLF is divided top-down to generate a candidate set of feasible functional groups.
  • Step 3 Module design: For each candidate function group, a module candidate proposal with computational functions and interfaces is created.
  • Step 4 Brain organ correspondence: From the candidate module group proposals, select those that can be mapped to existing brain organs.

To make Step 2 easier, we consider biological constraints as well as TLF specifications to narrow down the functional decomposition candidates. This allows us to generate BRA data for various brain regions using Function-Oriented SCID.

Each module obtained by this top-down approach has a defined function. Therefore, the BRA data can be refined by applying the usual SCID method to each module. In this presentation, we plan to illustrate the procedure of the proposed method using concrete examples such as imaginary architectures.

This work has been supported by the Mohammed bin Salman Center for Future Science and Technology for Saudi-Japan Vision 2030 at The University of Tokyo (MbSC2030).

  1. Yamakawa, H.  The whole brain architecture approach: Accelerating the development of artificial general intelligence by referring to the brain. Neural Netw. 2021, 144, 478-495.

Monday July 17, 2023 16:40 - 18:40 CEST

Attendees (6)