Histopathology-Based Deep-Learning Predicts Atherosclerotic Lesions in Intravascular Imaging
Background: Optical coherence tomography is a powerful modality to assess atherosclerotic lesions, but detecting lesions in high-resolution OCT is challenging and requires expert knowledge. Deep-learning algorithms can be used to automatically identify atherosclerotic lesions, facilitating identific...
| Main Authors: | , , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2021-12-01
|
| Series: | Frontiers in Cardiovascular Medicine |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fcvm.2021.779807/full |
| _version_ | 1831813384032485376 |
|---|---|
| author | Olle Holmberg Olle Holmberg Tobias Lenz Valentin Koch Valentin Koch Aseel Alyagoob Léa Utsch Andreas Rank Emina Sabic Masaru Seguchi Erion Xhepa Sebastian Kufner Salvatore Cassese Adnan Kastrati Adnan Kastrati Carsten Marr Carsten Marr Michael Joner Michael Joner Philipp Nicol |
| author_facet | Olle Holmberg Olle Holmberg Tobias Lenz Valentin Koch Valentin Koch Aseel Alyagoob Léa Utsch Andreas Rank Emina Sabic Masaru Seguchi Erion Xhepa Sebastian Kufner Salvatore Cassese Adnan Kastrati Adnan Kastrati Carsten Marr Carsten Marr Michael Joner Michael Joner Philipp Nicol |
| author_sort | Olle Holmberg |
| collection | DOAJ |
| description | Background: Optical coherence tomography is a powerful modality to assess atherosclerotic lesions, but detecting lesions in high-resolution OCT is challenging and requires expert knowledge. Deep-learning algorithms can be used to automatically identify atherosclerotic lesions, facilitating identification of patients at risk. We trained a deep-learning algorithm (DeepAD) with co-registered, annotated histopathology to predict atherosclerotic lesions in optical coherence tomography (OCT).Methods: Two datasets were used for training DeepAD: (i) a histopathology data set from 7 autopsy cases with 62 OCT frames and co-registered histopathology for high quality manual annotation and (ii) a clinical data set from 51 patients with 222 OCT frames in which manual annotations were based on clinical expertise only. A U-net based deep convolutional neural network (CNN) ensemble was employed as an atherosclerotic lesion prediction algorithm. Results were analyzed using intersection over union (IOU) for segmentation.Results: DeepAD showed good performance regarding the prediction of atherosclerotic lesions, with a median IOU of 0.68 ± 0.18 for segmentation of atherosclerotic lesions. Detection of calcified lesions yielded an IOU = 0.34. When training the algorithm without histopathology-based annotations, a performance drop of >0.25 IOU was observed. The practical application of DeepAD was evaluated retrospectively in a clinical cohort (n = 11 cases), showing high sensitivity as well as specificity and similar performance when compared to manual expert analysis.Conclusion: Automated detection of atherosclerotic lesions in OCT is improved using a histopathology-based deep-learning algorithm, allowing accurate detection in the clinical setting. An automated decision-support tool based on DeepAD could help in risk prediction and guide interventional treatment decisions. |
| first_indexed | 2024-12-22T21:53:16Z |
| format | Article |
| id | doaj.art-53d95aca30224c2dbe27d4e805522e43 |
| institution | Directory Open Access Journal |
| issn | 2297-055X |
| language | English |
| last_indexed | 2024-12-22T21:53:16Z |
| publishDate | 2021-12-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Cardiovascular Medicine |
| spelling | doaj.art-53d95aca30224c2dbe27d4e805522e432022-12-21T18:11:18ZengFrontiers Media S.A.Frontiers in Cardiovascular Medicine2297-055X2021-12-01810.3389/fcvm.2021.779807779807Histopathology-Based Deep-Learning Predicts Atherosclerotic Lesions in Intravascular ImagingOlle Holmberg0Olle Holmberg1Tobias Lenz2Valentin Koch3Valentin Koch4Aseel Alyagoob5Léa Utsch6Andreas Rank7Emina Sabic8Masaru Seguchi9Erion Xhepa10Sebastian Kufner11Salvatore Cassese12Adnan Kastrati13Adnan Kastrati14Carsten Marr15Carsten Marr16Michael Joner17Michael Joner18Philipp Nicol19Institute of Computational Biology, German Research Center for Environmental Health, Helmholtz Zentrum München, Oberschleißheim, GermanySchool of Life Sciences Weihenstephan, Technische Universität München, Munich, GermanyKlinik für Herz- und Kreislauferkrankungen, Deutsches Herzzentrum München, Technische Universität München, Munich, GermanyInstitute of AI for Health, German Research Center for Environmental Health, Helmholtz Zentrum München, Oberschleißheim, GermanyTUM Department of Informatics, Technische Universität München, Munich, GermanyKlinik für Herz- und Kreislauferkrankungen, Deutsches Herzzentrum München, Technische Universität München, Munich, GermanyKlinik für Herz- und Kreislauferkrankungen, Deutsches Herzzentrum München, Technische Universität München, Munich, GermanyKlinik für Herz- und Kreislauferkrankungen, Deutsches Herzzentrum München, Technische Universität München, Munich, GermanyKlinik für Herz- und Kreislauferkrankungen, Deutsches Herzzentrum München, Technische Universität München, Munich, GermanyKlinik für Herz- und Kreislauferkrankungen, Deutsches Herzzentrum München, Technische Universität München, Munich, GermanyKlinik für Herz- und Kreislauferkrankungen, Deutsches Herzzentrum München, Technische Universität München, Munich, GermanyKlinik für Herz- und Kreislauferkrankungen, Deutsches Herzzentrum München, Technische Universität München, Munich, GermanyKlinik für Herz- und Kreislauferkrankungen, Deutsches Herzzentrum München, Technische Universität München, Munich, GermanyKlinik für Herz- und Kreislauferkrankungen, Deutsches Herzzentrum München, Technische Universität München, Munich, GermanyDeutsches Zentrum für Herz- und Kreislauf-Forschung (DZHK) e.V. (German Center for Cardiovascular Research), Partner Site Munich Heart Alliance, Munich, GermanyInstitute of Computational Biology, German Research Center for Environmental Health, Helmholtz Zentrum München, Oberschleißheim, GermanyInstitute of AI for Health, German Research Center for Environmental Health, Helmholtz Zentrum München, Oberschleißheim, GermanyKlinik für Herz- und Kreislauferkrankungen, Deutsches Herzzentrum München, Technische Universität München, Munich, GermanyDeutsches Zentrum für Herz- und Kreislauf-Forschung (DZHK) e.V. (German Center for Cardiovascular Research), Partner Site Munich Heart Alliance, Munich, GermanyKlinik für Herz- und Kreislauferkrankungen, Deutsches Herzzentrum München, Technische Universität München, Munich, GermanyBackground: Optical coherence tomography is a powerful modality to assess atherosclerotic lesions, but detecting lesions in high-resolution OCT is challenging and requires expert knowledge. Deep-learning algorithms can be used to automatically identify atherosclerotic lesions, facilitating identification of patients at risk. We trained a deep-learning algorithm (DeepAD) with co-registered, annotated histopathology to predict atherosclerotic lesions in optical coherence tomography (OCT).Methods: Two datasets were used for training DeepAD: (i) a histopathology data set from 7 autopsy cases with 62 OCT frames and co-registered histopathology for high quality manual annotation and (ii) a clinical data set from 51 patients with 222 OCT frames in which manual annotations were based on clinical expertise only. A U-net based deep convolutional neural network (CNN) ensemble was employed as an atherosclerotic lesion prediction algorithm. Results were analyzed using intersection over union (IOU) for segmentation.Results: DeepAD showed good performance regarding the prediction of atherosclerotic lesions, with a median IOU of 0.68 ± 0.18 for segmentation of atherosclerotic lesions. Detection of calcified lesions yielded an IOU = 0.34. When training the algorithm without histopathology-based annotations, a performance drop of >0.25 IOU was observed. The practical application of DeepAD was evaluated retrospectively in a clinical cohort (n = 11 cases), showing high sensitivity as well as specificity and similar performance when compared to manual expert analysis.Conclusion: Automated detection of atherosclerotic lesions in OCT is improved using a histopathology-based deep-learning algorithm, allowing accurate detection in the clinical setting. An automated decision-support tool based on DeepAD could help in risk prediction and guide interventional treatment decisions.https://www.frontiersin.org/articles/10.3389/fcvm.2021.779807/fulldeep learningartificial intelligenceintravascular imagingatherosclerosishistopathologyoptical coherence tomography |
| spellingShingle | Olle Holmberg Olle Holmberg Tobias Lenz Valentin Koch Valentin Koch Aseel Alyagoob Léa Utsch Andreas Rank Emina Sabic Masaru Seguchi Erion Xhepa Sebastian Kufner Salvatore Cassese Adnan Kastrati Adnan Kastrati Carsten Marr Carsten Marr Michael Joner Michael Joner Philipp Nicol Histopathology-Based Deep-Learning Predicts Atherosclerotic Lesions in Intravascular Imaging Frontiers in Cardiovascular Medicine deep learning artificial intelligence intravascular imaging atherosclerosis histopathology optical coherence tomography |
| title | Histopathology-Based Deep-Learning Predicts Atherosclerotic Lesions in Intravascular Imaging |
| title_full | Histopathology-Based Deep-Learning Predicts Atherosclerotic Lesions in Intravascular Imaging |
| title_fullStr | Histopathology-Based Deep-Learning Predicts Atherosclerotic Lesions in Intravascular Imaging |
| title_full_unstemmed | Histopathology-Based Deep-Learning Predicts Atherosclerotic Lesions in Intravascular Imaging |
| title_short | Histopathology-Based Deep-Learning Predicts Atherosclerotic Lesions in Intravascular Imaging |
| title_sort | histopathology based deep learning predicts atherosclerotic lesions in intravascular imaging |
| topic | deep learning artificial intelligence intravascular imaging atherosclerosis histopathology optical coherence tomography |
| url | https://www.frontiersin.org/articles/10.3389/fcvm.2021.779807/full |
| work_keys_str_mv | AT olleholmberg histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT olleholmberg histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT tobiaslenz histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT valentinkoch histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT valentinkoch histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT aseelalyagoob histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT leautsch histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT andreasrank histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT eminasabic histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT masaruseguchi histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT erionxhepa histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT sebastiankufner histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT salvatorecassese histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT adnankastrati histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT adnankastrati histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT carstenmarr histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT carstenmarr histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT michaeljoner histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT michaeljoner histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging AT philippnicol histopathologybaseddeeplearningpredictsatheroscleroticlesionsinintravascularimaging |