{"id":{"repo_id":"cambridge","oai_identifier":"oai:www.repository.cam.ac.uk:1810/385075"},"canonical_url":"https://search.dev.ndltd.org/etd/cambridge/oai:www.repository.cam.ac.uk:1810/385075","repository":{"repo_id":"cambridge","name":"Cambridge University","base_url":"https://api.repository.cam.ac.uk/server/oai/request"},"display":{"title":"Advanced image processing methods for 3D high-throughput microscopy","abstract":"Advances in automated microscopy have led to the generation of vast and complex datasets, bringing bioimaging into the era of large data. This growth demands high-throughput image processing methods. This thesis addresses these challenges by developing efficient feature detection and 3D reconstruction methods, two key aspects in the image processing. For high-throughput feature detection, I developed a pipeline together with a new spot detection algorithm for large-scale 3D imaging of cells and tissues. It automates: (1) selecting in-focus images, (2) detecting features from labelled cells to diffraction-limited puncta, and (3) analysing spatial correlations between cells and proteins. Current tools lack the speed, precision and flexibility needed for large datasets, particularly for detecting puncta in complex environments. Together with neuroscientist collaborators, we have applied our new pipeline to process terabytes of image data, reducing analysis time from weeks to days. In doing so, we have revealed the spatial relationships between microglia, neurons, and α-synuclein oligomers in human brain tissue. For high-throughput 3D reconstruction, I proposed a fast reconstruction algorithm for Fourier light-field microscopy (FLFM). (FLFM) captures 3D information in a single 2D frame, thus requiring a reconstruction algorithm to recover the original 3D volume. Traditional FLFM reconstruction methods are computationally intensive, achieving speeds of only 0.05–0.1 cells per second—too slow for applications such as fluorescence-activated cell sorting (FACS). Therefore, a new reconstruction method, named ‘patch deconvolution’, was proposed, which accelerates reconstruction to 5 cells per second on standard CPUs — a 50 fold improvement. Validated on simulated and experimental data, it achieves comparable accuracy to current methods and potentially enables real-time sorting based on 3D features. In summary, the proposed methods advance the analysis of complex datasets and support new biological applications, including understanding α-synuclein oligomers in Parkinson’s Disease and sorting cells in real-time based on spatial features.","abstract_html":"Advances in automated microscopy have led to the generation of vast and complex datasets, bringing bioimaging into the era of large data. This growth demands high-throughput image processing methods. This thesis addresses these challenges by developing efficient feature detection and 3D reconstruction methods, two key aspects in the image processing. For high-throughput feature detection, I developed a pipeline together with a new spot detection algorithm for large-scale 3D imaging of cells and tissues. It automates: (1) selecting in-focus images, (2) detecting features from labelled cells to diffraction-limited puncta, and (3) analysing spatial correlations between cells and proteins. Current tools lack the speed, precision and flexibility needed for large datasets, particularly for detecting puncta in complex environments. Together with neuroscientist collaborators, we have applied our new pipeline to process terabytes of image data, reducing analysis time from weeks to days. In doing so, we have revealed the spatial relationships between microglia, neurons, and α-synuclein oligomers in human brain tissue. For high-throughput 3D reconstruction, I proposed a fast reconstruction algorithm for Fourier light-field microscopy (FLFM). (FLFM) captures 3D information in a single 2D frame, thus requiring a reconstruction algorithm to recover the original 3D volume. Traditional FLFM reconstruction methods are computationally intensive, achieving speeds of only 0.05–0.1 cells per second—too slow for applications such as fluorescence-activated cell sorting (FACS). Therefore, a new reconstruction method, named ‘patch deconvolution’, was proposed, which accelerates reconstruction to 5 cells per second on standard CPUs — a 50 fold improvement. Validated on simulated and experimental data, it achieves comparable accuracy to current methods and potentially enables real-time sorting based on 3D features. In summary, the proposed methods advance the analysis of complex datasets and support new biological applications, including understanding α-synuclein oligomers in Parkinson’s Disease and sorting cells in real-time based on spatial features.","abstract_has_math":false,"creators":["Fu, Bin"],"institution":"University of Cambridge","degree_name":"Doctor of Philosophy (PhD)","degree_level":"Doctoral","degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":["Lee, Steven"],"committee_chairs":[],"committee_members":[],"year":2025,"date_issued":"2025-03-14","date_published":"2025-03-14","updated_at":"2026-07-22T22:24:30Z","subjects":["Deconvolution","Flow cytometry","Image processing","Microscopy"],"languages":[],"rights":[],"rights_urls":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/5eab9909-8f6c-4591-b38c-ad4a76a158f3/download","http://purl.org/NET/rdflicense/allrightsreserved"],"identifier_entries":[{"key":"dc:creator.authoridentifier","label":"Author Identifier","values":["0000000288162906"],"render_values":[{"text":"0000-0002-8816-2906","href":"https://orcid.org/0000-0002-8816-2906","code":true}]}]},"links":{"outbound_url":"https://doi.org/10.17863/CAM.118852","outbound_label":"DOI","outbound_source":"dc:identifier.doi"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Lee, Steven"]},{"key":"dc:contributor.sponsor","label":"Sponsor","values":["Aligning Science Across Parkinson's [ASAP-000509] through the Michael J. Fox Foundation for Parkinsons Research (MJFF)."]},{"key":"dc:creator","label":"Author","values":["Fu, Bin"]},{"key":"dc:creator.authoridentifier","label":"Author Identifier","values":["0000000288162906"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.issued","label":"Date","values":["2025-03-14"]},{"key":"dc:publisher.institution","label":"Dc Publisher Institution","values":["University of Cambridge"]},{"key":"dc:relation.isreferencedby.uri","label":"Dc Relation Isreferencedby URI","values":["https://www.repository.cam.ac.uk/handle/1810/385075"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"dc:type.qualificationlevel","label":"Dc Type Qualificationlevel","values":["Doctoral"]},{"key":"dc:type.qualificationname","label":"Dc Type Qualificationname","values":["Doctor of Philosophy (PhD)"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Deconvolution","Flow cytometry","Image processing","Microscopy"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:rights","label":"Dc Rights","values":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/5eab9909-8f6c-4591-b38c-ad4a76a158f3/download","http://purl.org/NET/rdflicense/allrightsreserved"]},{"key":"dc:rights.embargodate","label":"Dc Rights Embargodate","values":["2026-06-04"]},{"key":"dc:rights.embargotype","label":"Dc Rights Embargotype","values":["embargo"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.doi","label":"DOI","values":["https://doi.org/10.17863/CAM.118852"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/04127c07-5521-4236-8141-4c13a306c1f4/download"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Advances in automated microscopy have led to the generation of vast and complex datasets, bringing bioimaging into the era of large data. This growth demands high-throughput image processing methods. This thesis addresses these challenges by developing efficient feature detection and 3D reconstruction methods, two key aspects in the image processing. For high-throughput feature detection, I developed a pipeline together with a new spot detection algorithm for large-scale 3D imaging of cells and tissues. It automates: (1) selecting in-focus images, (2) detecting features from labelled cells to diffraction-limited puncta, and (3) analysing spatial correlations between cells and proteins. Current tools lack the speed, precision and flexibility needed for large datasets, particularly for detecting puncta in complex environments. Together with neuroscientist collaborators, we have applied our new pipeline to process terabytes of image data, reducing analysis time from weeks to days. In doing so, we have revealed the spatial relationships between microglia, neurons, and α-synuclein oligomers in human brain tissue. For high-throughput 3D reconstruction, I proposed a fast reconstruction algorithm for Fourier light-field microscopy (FLFM). (FLFM) captures 3D information in a single 2D frame, thus requiring a reconstruction algorithm to recover the original 3D volume. Traditional FLFM reconstruction methods are computationally intensive, achieving speeds of only 0.05–0.1 cells per second—too slow for applications such as fluorescence-activated cell sorting (FACS). Therefore, a new reconstruction method, named ‘patch deconvolution’, was proposed, which accelerates reconstruction to 5 cells per second on standard CPUs — a 50 fold improvement. Validated on simulated and experimental data, it achieves comparable accuracy to current methods and potentially enables real-time sorting based on 3D features. In summary, the proposed methods advance the analysis of complex datasets and support new biological applications, including understanding α-synuclein oligomers in Parkinson’s Disease and sorting cells in real-time based on spatial features."]},{"key":"dc:format.checksum.md5","label":"Dc Format Checksum Md5","values":["25a4dd9c9ab273a4cd89553da85d0231","87eda9de84448d1f82354d60eee3eb5f"]},{"key":"dc:title","label":"Title","values":["Advanced image processing methods for 3D high-throughput microscopy"]}]}],"canonical_facts":{"dc:contributor.advisor":["Lee, Steven"],"dc:contributor.sponsor":["Aligning Science Across Parkinson's [ASAP-000509] through the Michael J. Fox Foundation for Parkinsons Research (MJFF)."],"dc:creator":["Fu, Bin"],"dc:creator.authoridentifier":["0000000288162906"],"dc:date.issued":["2025-03-14"],"dc:description.abstract":["Advances in automated microscopy have led to the generation of vast and complex datasets, bringing bioimaging into the era of large data. This growth demands high-throughput image processing methods. This thesis addresses these challenges by developing efficient feature detection and 3D reconstruction methods, two key aspects in the image processing. For high-throughput feature detection, I developed a pipeline together with a new spot detection algorithm for large-scale 3D imaging of cells and tissues. It automates: (1) selecting in-focus images, (2) detecting features from labelled cells to diffraction-limited puncta, and (3) analysing spatial correlations between cells and proteins. Current tools lack the speed, precision and flexibility needed for large datasets, particularly for detecting puncta in complex environments. Together with neuroscientist collaborators, we have applied our new pipeline to process terabytes of image data, reducing analysis time from weeks to days. In doing so, we have revealed the spatial relationships between microglia, neurons, and α-synuclein oligomers in human brain tissue. For high-throughput 3D reconstruction, I proposed a fast reconstruction algorithm for Fourier light-field microscopy (FLFM). (FLFM) captures 3D information in a single 2D frame, thus requiring a reconstruction algorithm to recover the original 3D volume. Traditional FLFM reconstruction methods are computationally intensive, achieving speeds of only 0.05–0.1 cells per second—too slow for applications such as fluorescence-activated cell sorting (FACS). Therefore, a new reconstruction method, named ‘patch deconvolution’, was proposed, which accelerates reconstruction to 5 cells per second on standard CPUs — a 50 fold improvement. Validated on simulated and experimental data, it achieves comparable accuracy to current methods and potentially enables real-time sorting based on 3D features. In summary, the proposed methods advance the analysis of complex datasets and support new biological applications, including understanding α-synuclein oligomers in Parkinson’s Disease and sorting cells in real-time based on spatial features."],"dc:format.checksum.md5":["25a4dd9c9ab273a4cd89553da85d0231","87eda9de84448d1f82354d60eee3eb5f"],"dc:identifier.doi":["https://doi.org/10.17863/CAM.118852"],"dc:identifier.uri":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/04127c07-5521-4236-8141-4c13a306c1f4/download"],"dc:publisher.institution":["University of Cambridge"],"dc:relation.isreferencedby.uri":["https://www.repository.cam.ac.uk/handle/1810/385075"],"dc:rights":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/5eab9909-8f6c-4591-b38c-ad4a76a158f3/download","http://purl.org/NET/rdflicense/allrightsreserved"],"dc:rights.embargodate":["2026-06-04"],"dc:rights.embargotype":["embargo"],"dc:subject":["Deconvolution","Flow cytometry","Image processing","Microscopy"],"dc:title":["Advanced image processing methods for 3D high-throughput microscopy"],"dc:type":["Thesis"],"dc:type.qualificationlevel":["Doctoral"],"dc:type.qualificationname":["Doctor of Philosophy (PhD)"]},"updated_at":"2026-07-22T22:24:30Z"}