The energy metabolic footprint of predictive processing in the human brain

Hechler, André (2024): The energy metabolic footprint of predictive processing in the human brain. Dissertation, LMU München: Graduate School of Systemic Neurosciences (GSN)

Abstract

To maximize our chances of survival and procreation, we need to process our environment in a highly sophisticated and accurate manner. In their limit, these two demands are mutually exclusive: While better sound localization, quicker reflexes or more accurate vision could improve survivability, the necessary energy consumption might not be sustainable. Luckily, our sensory systems strike an impressive balance between performance and energetic cost. In a both active and passive process, we learn about the rules that determine our experience and use them to form expectations. Efficient brain activity is then achieved by limiting the forward transmission of signals to deviations from what we predicted. In the visual domain, this means that our perception is dominated by our expectations when we are in a familiar environment. Research in cognitive neuroscience has shown that expected input elicits weaker brain activity than surprising input, without any behavioral disadvantages. However, knowledge about associated energetic efficiency is limited by three gaps in the current literature. First, conventional imaging techniques do not provide direct measurements of energy metabolism. Second, previous research has focused on localizing areas of maximal effect, potentially missing weaker, but more widespread patterns. Third, our knowledge about the world is imperfect, leading to uncertain expectations. This has rarely been accounted for. Neuronal activity is fueled by ATP, most of which is produced with chemical reactions that need oxygen. In the present work, I assessed energy metabolism with a novel imaging method that measures the rate of oxygen consumption across all parts of the brain. I used an experimental design during which participants saw visual object sequences that were either predictable, random, or surprising. Behavioral tests indicated that predictable sequences were learned without any feedback which resulted in anticipation of upcoming objects. I further found that participants varied in the confidence of their expectations. This had a major impact on oxygen consumption when viewing predictable sequences: The lowest energy usage was found for high levels of confidence. This effect was not limited to sensory regions but extended across large parts of the brain. Interestingly, my results suggest that confidence led to energy savings even when the visual input was objectively random. In conclusion, this work provides the first evidence that our expectations are a major promoter of efficient processing, which is crucial for any organism with limited energy availability.

• Dokumententyp: Dissertationen (Dissertation, LMU München)
• Keywords: Neuroscience, Cognition, Imaging, Prediction Efficiency
• Themengebiete: 500 Naturwissenschaften und Mathematik ; 500 Naturwissenschaften und Mathematik > 570 Biowissenschaften, Biologie
• Fakultäten: Graduate School of Systemic Neurosciences (GSN)
• Sprache der Hochschulschrift: Englisch
• Datum der mündlichen Prüfung: 14. Mai 2024
• 1. Bericht­erstatter:in: Riedl, Valentin