Evolutionary design of explainable algorithms for biomedical image segmentation
Abstract An unresolved issue in contemporary biomedicine is the overwhelming number and diversity of complex images that require annotation, analysis and interpretation. Recent advances in Deep Learning have revolutionized the field of computer vision, creating algorithms that compete with human exp...
| Main Authors: | , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2023-11-01
|
| Series: | Nature Communications |
| Online Access: | https://doi.org/10.1038/s41467-023-42664-x |
| _version_ | 1827764291170729984 |
|---|---|
| author | Kévin Cortacero Brienne McKenzie Sabina Müller Roxana Khazen Fanny Lafouresse Gaëlle Corsaut Nathalie Van Acker François-Xavier Frenois Laurence Lamant Nicolas Meyer Béatrice Vergier Dennis G. Wilson Hervé Luga Oskar Staufer Michael L. Dustin Salvatore Valitutti Sylvain Cussat-Blanc |
| author_facet | Kévin Cortacero Brienne McKenzie Sabina Müller Roxana Khazen Fanny Lafouresse Gaëlle Corsaut Nathalie Van Acker François-Xavier Frenois Laurence Lamant Nicolas Meyer Béatrice Vergier Dennis G. Wilson Hervé Luga Oskar Staufer Michael L. Dustin Salvatore Valitutti Sylvain Cussat-Blanc |
| author_sort | Kévin Cortacero |
| collection | DOAJ |
| description | Abstract An unresolved issue in contemporary biomedicine is the overwhelming number and diversity of complex images that require annotation, analysis and interpretation. Recent advances in Deep Learning have revolutionized the field of computer vision, creating algorithms that compete with human experts in image segmentation tasks. However, these frameworks require large human-annotated datasets for training and the resulting “black box” models are difficult to interpret. In this study, we introduce Kartezio, a modular Cartesian Genetic Programming-based computational strategy that generates fully transparent and easily interpretable image processing pipelines by iteratively assembling and parameterizing computer vision functions. The pipelines thus generated exhibit comparable precision to state-of-the-art Deep Learning approaches on instance segmentation tasks, while requiring drastically smaller training datasets. This Few-Shot Learning method confers tremendous flexibility, speed, and functionality to this approach. We then deploy Kartezio to solve a series of semantic and instance segmentation problems, and demonstrate its utility across diverse images ranging from multiplexed tissue histopathology images to high resolution microscopy images. While the flexibility, robustness and practical utility of Kartezio make this fully explicable evolutionary designer a potential game-changer in the field of biomedical image processing, Kartezio remains complementary and potentially auxiliary to mainstream Deep Learning approaches. |
| first_indexed | 2024-03-11T11:03:18Z |
| format | Article |
| id | doaj.art-755405e7d248417ea16e00b6baf15051 |
| institution | Directory Open Access Journal |
| issn | 2041-1723 |
| language | English |
| last_indexed | 2024-03-11T11:03:18Z |
| publishDate | 2023-11-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Nature Communications |
| spelling | doaj.art-755405e7d248417ea16e00b6baf150512023-11-12T12:22:44ZengNature PortfolioNature Communications2041-17232023-11-0114111810.1038/s41467-023-42664-xEvolutionary design of explainable algorithms for biomedical image segmentationKévin Cortacero0Brienne McKenzie1Sabina Müller2Roxana Khazen3Fanny Lafouresse4Gaëlle Corsaut5Nathalie Van Acker6François-Xavier Frenois7Laurence Lamant8Nicolas Meyer9Béatrice Vergier10Dennis G. Wilson11Hervé Luga12Oskar Staufer13Michael L. Dustin14Salvatore Valitutti15Sylvain Cussat-Blanc16Institut National de la Santé et de la Recherche Médicale (INSERM) UMR1037, Centre de Recherche en Cancérologie de Toulouse (CRCT)Institut National de la Santé et de la Recherche Médicale (INSERM) UMR1037, Centre de Recherche en Cancérologie de Toulouse (CRCT)Institut National de la Santé et de la Recherche Médicale (INSERM) UMR1037, Centre de Recherche en Cancérologie de Toulouse (CRCT)Institut National de la Santé et de la Recherche Médicale (INSERM) UMR1037, Centre de Recherche en Cancérologie de Toulouse (CRCT)Institut National de la Santé et de la Recherche Médicale (INSERM) UMR1037, Centre de Recherche en Cancérologie de Toulouse (CRCT)Institut National de la Santé et de la Recherche Médicale (INSERM) UMR1037, Centre de Recherche en Cancérologie de Toulouse (CRCT)Department of Pathology, Institut Universitaire du Cancer-Oncopole de Toulouse (IUCT)Department of Pathology, Institut Universitaire du Cancer-Oncopole de Toulouse (IUCT)Department of Pathology, Institut Universitaire du Cancer-Oncopole de Toulouse (IUCT)Department of Dermatology, IUCTService de Pathologie, Centre Hospitalier Universitaire de BordeauxUniversity of Toulouse - Institut de Recherche en Informatique de Toulouse (IRIT) - UMR5505, Artificial and Natural Intelligence Toulouse InstituteUniversity of Toulouse - Institut de Recherche en Informatique de Toulouse (IRIT) - UMR5505, Artificial and Natural Intelligence Toulouse InstituteKennedy Institute of Rheumatology (KIR), Nuffield Department of Orthopedics, Rheumatology and Musculoskeletal Sciences, University of OxfordKennedy Institute of Rheumatology (KIR), Nuffield Department of Orthopedics, Rheumatology and Musculoskeletal Sciences, University of OxfordInstitut National de la Santé et de la Recherche Médicale (INSERM) UMR1037, Centre de Recherche en Cancérologie de Toulouse (CRCT)University of Toulouse - Institut de Recherche en Informatique de Toulouse (IRIT) - UMR5505, Artificial and Natural Intelligence Toulouse InstituteAbstract An unresolved issue in contemporary biomedicine is the overwhelming number and diversity of complex images that require annotation, analysis and interpretation. Recent advances in Deep Learning have revolutionized the field of computer vision, creating algorithms that compete with human experts in image segmentation tasks. However, these frameworks require large human-annotated datasets for training and the resulting “black box” models are difficult to interpret. In this study, we introduce Kartezio, a modular Cartesian Genetic Programming-based computational strategy that generates fully transparent and easily interpretable image processing pipelines by iteratively assembling and parameterizing computer vision functions. The pipelines thus generated exhibit comparable precision to state-of-the-art Deep Learning approaches on instance segmentation tasks, while requiring drastically smaller training datasets. This Few-Shot Learning method confers tremendous flexibility, speed, and functionality to this approach. We then deploy Kartezio to solve a series of semantic and instance segmentation problems, and demonstrate its utility across diverse images ranging from multiplexed tissue histopathology images to high resolution microscopy images. While the flexibility, robustness and practical utility of Kartezio make this fully explicable evolutionary designer a potential game-changer in the field of biomedical image processing, Kartezio remains complementary and potentially auxiliary to mainstream Deep Learning approaches.https://doi.org/10.1038/s41467-023-42664-x |
| spellingShingle | Kévin Cortacero Brienne McKenzie Sabina Müller Roxana Khazen Fanny Lafouresse Gaëlle Corsaut Nathalie Van Acker François-Xavier Frenois Laurence Lamant Nicolas Meyer Béatrice Vergier Dennis G. Wilson Hervé Luga Oskar Staufer Michael L. Dustin Salvatore Valitutti Sylvain Cussat-Blanc Evolutionary design of explainable algorithms for biomedical image segmentation Nature Communications |
| title | Evolutionary design of explainable algorithms for biomedical image segmentation |
| title_full | Evolutionary design of explainable algorithms for biomedical image segmentation |
| title_fullStr | Evolutionary design of explainable algorithms for biomedical image segmentation |
| title_full_unstemmed | Evolutionary design of explainable algorithms for biomedical image segmentation |
| title_short | Evolutionary design of explainable algorithms for biomedical image segmentation |
| title_sort | evolutionary design of explainable algorithms for biomedical image segmentation |
| url | https://doi.org/10.1038/s41467-023-42664-x |
| work_keys_str_mv | AT kevincortacero evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation AT briennemckenzie evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation AT sabinamuller evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation AT roxanakhazen evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation AT fannylafouresse evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation AT gaellecorsaut evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation AT nathalievanacker evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation AT francoisxavierfrenois evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation AT laurencelamant evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation AT nicolasmeyer evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation AT beatricevergier evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation AT dennisgwilson evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation AT herveluga evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation AT oskarstaufer evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation AT michaelldustin evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation AT salvatorevalitutti evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation AT sylvaincussatblanc evolutionarydesignofexplainablealgorithmsforbiomedicalimagesegmentation |