Pushing the boundaries of single molecule microscopy, optical sensing and mass spectrometry through deep learning
Doctoral thesis, 2024
Deep learning has transformed sensing and characterization technologies, enabling significant advancements across various scientific domains. This thesis investigates the application of deep learning techniques to enhance the applicability of microscopy, spectrometry, and sensing, particularly under high noise conditions. The central hypothesis of this research is that deep learning can substantially improve the sensitivity and specificity of these technologies, allowing for the detection and analysis of minute signals that were previously obscured by the noise.
Key contributions include the development of novel deep learning methods that enhance nanofluidic scattering microscopy, nanoplasmonic sensing, and mass spectrometry. These methods enable precise quantification of chemical reactions on nanoscale surfaces, detailed detection of cellular structures and molecular interactions, and accurate identification of low-concentration substances amidst strong background signals. Collectively, these advancements push the boundaries of what can be measured and observed at microscopic and molecular scales, offering groundbreaking applications in environmental monitoring and healthcare diagnostics.
Key contributions include the development of novel deep learning methods that enhance nanofluidic scattering microscopy, nanoplasmonic sensing, and mass spectrometry. These methods enable precise quantification of chemical reactions on nanoscale surfaces, detailed detection of cellular structures and molecular interactions, and accurate identification of low-concentration substances amidst strong background signals. Collectively, these advancements push the boundaries of what can be measured and observed at microscopic and molecular scales, offering groundbreaking applications in environmental monitoring and healthcare diagnostics.
Mass Spectrometry
Nanoplasmonic Sensing
Environmental Monitoring
Virtual Staining
Cross-modality Transformation
Nanofluidic Scattering Microscopy
Deep Learning
Biomolecule Detection
Author
Henrik Klein Moberg
Chalmers, Physics, Chemical Physics
Included papers
Label-free nanofluidic scattering microscopy of size and mass of single diffusing molecules and nanoparticles
Nature Methods,;Vol. 19(2022)p. 751-758
Journal article
Neural network enabled nanoplasmonic hydrogen sensors with 100 ppm limit of detection in humid air
Nature Communications,;Vol. 15(2024)
Journal article
Henrik Klein Moberg, Joachim Fritzsche, Giovanni Volpe, Christoph Langhammer, Deep-learning enabled mass spectrometry of the reaction product from a single catalyst nanoparticle
Manuscript
Viktor Martvall, Henrik Klein Moberg, Athanasios Theodiridis, David Tomecek, Pernilla Tanner, Paul Erhart, Christoph Langhammer, Accelerating Plasmonic Hydrogen Sensors for Inert Gas Environments by Transformer-Based Deep Learning
Manuscript
Cross-modality transformations in biological microscopy enabled by deep learning
Advanced Photonics,;Vol. 6(2024)
Review article
Henrik Klein Moberg, Bohdan Yeroshenko, David Albinsson, Joachim Fritzsche, Daniel Midtvedt, Barbora Spackova, Giovanni Volpe, Christoph Langhammer, Deep Learning Microscopy for Label-free Biomolecule Weight-and-Size Characterization in the Single-kDa Regime
Manuscript
Popular science description
English
This PhD thesis focuses on how artificial intelligence (AI) is being used to help scientists study molecules and chemical reactions in a new and more powerful way. Scientists use tools like microscopes and sensors to look at tiny molecules, but these tools have limitations, especially when dealing with very small molecules or when there is a lot of noise, making it hard to get clear results. This research introduces AI to overcome these challenges by recognizing patterns that human eyes or traditional tools might miss.
One key breakthrough in this work is improving a special type of microscope, called Nanofluidic Scattering Microscopy (NSM), which lets scientists observe molecules like insulin without having to add any labels or dyes. This allows the study of molecules in their natural state, making the results more accurate. The AI also makes it possible to detect these tiny molecules in real-time, even when they are hard to see with traditional methods.
AI is also being used to make hydrogen sensors more precise. These sensors are important for the safe use of hydrogen as a clean energy source. By using AI, the sensors can detect hydrogen more accurately, which is crucial for safety in industries using hydrogen fuel.
Overall, this thesis shows how AI is helping scientists push the limits of what we can see and measure in the world of tiny molecules. This technology can have big impacts on healthcare, energy, and the environment by making it easier to study and understand the small details that matter.
One key breakthrough in this work is improving a special type of microscope, called Nanofluidic Scattering Microscopy (NSM), which lets scientists observe molecules like insulin without having to add any labels or dyes. This allows the study of molecules in their natural state, making the results more accurate. The AI also makes it possible to detect these tiny molecules in real-time, even when they are hard to see with traditional methods.
AI is also being used to make hydrogen sensors more precise. These sensors are important for the safe use of hydrogen as a clean energy source. By using AI, the sensors can detect hydrogen more accurately, which is crucial for safety in industries using hydrogen fuel.
Overall, this thesis shows how AI is helping scientists push the limits of what we can see and measure in the world of tiny molecules. This technology can have big impacts on healthcare, energy, and the environment by making it easier to study and understand the small details that matter.
Research Project(s)
NACAREI: Nanofluidic Catalytic Reaction Imaging
European Commission (EC) (101043480), 2023-01-01 -- 2027-12-31.
The Sub-10 nm Challenge in Single Particle Catalysis
Swedish Research Council (VR) (2018-00329), 2019-01-01 -- 2024-12-31.
Nanoplasmonisk Sensing's Heliga Graal
Swedish Foundation for Strategic Research (SSF) (FFL15-0087), 2017-01-01 -- 2021-12-31.
Categorizing
Areas of Advance
Nanoscience and Nanotechnology
Infrastructure
C3SE (-2020, Chalmers Centre for Computational Science and Engineering)
Chalmers Materials Analysis Laboratory
Myfab (incl. Nanofabrication Laboratory)
Subject Categories (SSIF 2011)
Other Physics Topics
Nano Technology
Computer Science
Identifiers
ISBN
978-91-8103-118-8
Other
Series
Doktorsavhandlingar vid Chalmers tekniska högskola. Ny serie: 5576
Publisher
Chalmers
Public defence
2024-11-13 09:30 -- 12:00
PJ Fysik Origo, Fysikgården 4
Opponent: Prof. Dr. Ir. Fons J Verbeek, Inst. of Advanced Computer Science, Leiden University