Deep learning for the detection of anatomical tissue structures and neoplasms of the skin on scanned histopathological tissue sections
Basal cell carcinoma (BCC), squamous cell carcinoma (SqCC) and melanoma are among the most common cancer types. Correct diagnosis based on histological evaluation after biopsy or excision is paramount for adequate therapy stratification. Deep learning on histological slides has been suggested to com...
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2022-11-01
|
| Series: | Frontiers in Oncology |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fonc.2022.1022967/full |
| _version_ | 1797824044368658432 |
|---|---|
| author | Katharina Kriegsmann Frithjof Lobers Christiane Zgorzelski Jörg Kriegsmann Jörg Kriegsmann Charlotte Janßen Rolf Rüdinger Meliß Thomas Muley Ulrich Sack Georg Steinbuss Mark Kriegsmann |
| author_facet | Katharina Kriegsmann Frithjof Lobers Christiane Zgorzelski Jörg Kriegsmann Jörg Kriegsmann Charlotte Janßen Rolf Rüdinger Meliß Thomas Muley Ulrich Sack Georg Steinbuss Mark Kriegsmann |
| author_sort | Katharina Kriegsmann |
| collection | DOAJ |
| description | Basal cell carcinoma (BCC), squamous cell carcinoma (SqCC) and melanoma are among the most common cancer types. Correct diagnosis based on histological evaluation after biopsy or excision is paramount for adequate therapy stratification. Deep learning on histological slides has been suggested to complement and improve routine diagnostics, but publicly available curated and annotated data and usable models trained to distinguish common skin tumors are rare and often lack heterogeneous non-tumor categories. A total of 16 classes from 386 cases were manually annotated on scanned histological slides, 129,364 100 x 100 µm (~395 x 395 px) image tiles were extracted and split into a training, validation and test set. An EfficientV2 neuronal network was trained and optimized to classify image categories. Cross entropy loss, balanced accuracy and Matthews correlation coefficient were used for model evaluation. Image and patient data were assessed with confusion matrices. Application of the model to an external set of whole slides facilitated localization of melanoma and non-tumor tissue. Automated differentiation of BCC, SqCC, melanoma, naevi and non-tumor tissue structures was possible, and a high diagnostic accuracy was achieved in the validation (98%) and test (97%) set. In summary, we provide a curated dataset including the most common neoplasms of the skin and various anatomical compartments to enable researchers to train, validate and improve deep learning models. Automated classification of skin tumors by deep learning techniques is possible with high accuracy, facilitates tumor localization and has the potential to support and improve routine diagnostics. |
| first_indexed | 2024-03-13T10:33:08Z |
| format | Article |
| id | doaj.art-8840ad37a52c49008f9302c6b2c2d6e3 |
| institution | Directory Open Access Journal |
| issn | 2234-943X |
| language | English |
| last_indexed | 2024-03-13T10:33:08Z |
| publishDate | 2022-11-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Oncology |
| spelling | doaj.art-8840ad37a52c49008f9302c6b2c2d6e32023-05-18T08:17:42ZengFrontiers Media S.A.Frontiers in Oncology2234-943X2022-11-011210.3389/fonc.2022.10229671022967Deep learning for the detection of anatomical tissue structures and neoplasms of the skin on scanned histopathological tissue sectionsKatharina Kriegsmann0Frithjof Lobers1Christiane Zgorzelski2Jörg Kriegsmann3Jörg Kriegsmann4Charlotte Janßen5Rolf Rüdinger Meliß6Thomas Muley7Ulrich Sack8Georg Steinbuss9Mark Kriegsmann10Department of Hematology, Oncology and Rheumatology, Heidelberg University, Heidelberg, GermanyDepartment of Clinical Immunology, Medical Faculty, University of Leipzig, Leipzig, GermanyInstitute of Pathology, Heidelberg University, Heidelberg, GermanyMVZ Histology, Cytology and Molecular Diagnostics Trier, Trier, GermanyProteopath Trier, Trier, GermanyCenter for Industrial Mathematics (ZeTeM), University of Bremen, Bremen, GermanyInstitute for Dermatopathology, Hannover, GermanyTranslational Lung Research Centre (TLRC) Heidelberg, Member of the German Centre for Lung Research (DZL), Heidelberg, GermanyDepartment of Clinical Immunology, Medical Faculty, University of Leipzig, Leipzig, GermanyDepartment of Hematology, Oncology and Rheumatology, Heidelberg University, Heidelberg, GermanyInstitute of Pathology, Heidelberg University, Heidelberg, GermanyBasal cell carcinoma (BCC), squamous cell carcinoma (SqCC) and melanoma are among the most common cancer types. Correct diagnosis based on histological evaluation after biopsy or excision is paramount for adequate therapy stratification. Deep learning on histological slides has been suggested to complement and improve routine diagnostics, but publicly available curated and annotated data and usable models trained to distinguish common skin tumors are rare and often lack heterogeneous non-tumor categories. A total of 16 classes from 386 cases were manually annotated on scanned histological slides, 129,364 100 x 100 µm (~395 x 395 px) image tiles were extracted and split into a training, validation and test set. An EfficientV2 neuronal network was trained and optimized to classify image categories. Cross entropy loss, balanced accuracy and Matthews correlation coefficient were used for model evaluation. Image and patient data were assessed with confusion matrices. Application of the model to an external set of whole slides facilitated localization of melanoma and non-tumor tissue. Automated differentiation of BCC, SqCC, melanoma, naevi and non-tumor tissue structures was possible, and a high diagnostic accuracy was achieved in the validation (98%) and test (97%) set. In summary, we provide a curated dataset including the most common neoplasms of the skin and various anatomical compartments to enable researchers to train, validate and improve deep learning models. Automated classification of skin tumors by deep learning techniques is possible with high accuracy, facilitates tumor localization and has the potential to support and improve routine diagnostics.https://www.frontiersin.org/articles/10.3389/fonc.2022.1022967/fulldeep learningpathologyartificial intelligencedermatopathologydigital pathologydeep learning - artificial neural network |
| spellingShingle | Katharina Kriegsmann Frithjof Lobers Christiane Zgorzelski Jörg Kriegsmann Jörg Kriegsmann Charlotte Janßen Rolf Rüdinger Meliß Thomas Muley Ulrich Sack Georg Steinbuss Mark Kriegsmann Deep learning for the detection of anatomical tissue structures and neoplasms of the skin on scanned histopathological tissue sections Frontiers in Oncology deep learning pathology artificial intelligence dermatopathology digital pathology deep learning - artificial neural network |
| title | Deep learning for the detection of anatomical tissue structures and neoplasms of the skin on scanned histopathological tissue sections |
| title_full | Deep learning for the detection of anatomical tissue structures and neoplasms of the skin on scanned histopathological tissue sections |
| title_fullStr | Deep learning for the detection of anatomical tissue structures and neoplasms of the skin on scanned histopathological tissue sections |
| title_full_unstemmed | Deep learning for the detection of anatomical tissue structures and neoplasms of the skin on scanned histopathological tissue sections |
| title_short | Deep learning for the detection of anatomical tissue structures and neoplasms of the skin on scanned histopathological tissue sections |
| title_sort | deep learning for the detection of anatomical tissue structures and neoplasms of the skin on scanned histopathological tissue sections |
| topic | deep learning pathology artificial intelligence dermatopathology digital pathology deep learning - artificial neural network |
| url | https://www.frontiersin.org/articles/10.3389/fonc.2022.1022967/full |
| work_keys_str_mv | AT katharinakriegsmann deeplearningforthedetectionofanatomicaltissuestructuresandneoplasmsoftheskinonscannedhistopathologicaltissuesections AT frithjoflobers deeplearningforthedetectionofanatomicaltissuestructuresandneoplasmsoftheskinonscannedhistopathologicaltissuesections AT christianezgorzelski deeplearningforthedetectionofanatomicaltissuestructuresandneoplasmsoftheskinonscannedhistopathologicaltissuesections AT jorgkriegsmann deeplearningforthedetectionofanatomicaltissuestructuresandneoplasmsoftheskinonscannedhistopathologicaltissuesections AT jorgkriegsmann deeplearningforthedetectionofanatomicaltissuestructuresandneoplasmsoftheskinonscannedhistopathologicaltissuesections AT charlottejanßen deeplearningforthedetectionofanatomicaltissuestructuresandneoplasmsoftheskinonscannedhistopathologicaltissuesections AT rolfrudingermeliß deeplearningforthedetectionofanatomicaltissuestructuresandneoplasmsoftheskinonscannedhistopathologicaltissuesections AT thomasmuley deeplearningforthedetectionofanatomicaltissuestructuresandneoplasmsoftheskinonscannedhistopathologicaltissuesections AT ulrichsack deeplearningforthedetectionofanatomicaltissuestructuresandneoplasmsoftheskinonscannedhistopathologicaltissuesections AT georgsteinbuss deeplearningforthedetectionofanatomicaltissuestructuresandneoplasmsoftheskinonscannedhistopathologicaltissuesections AT markkriegsmann deeplearningforthedetectionofanatomicaltissuestructuresandneoplasmsoftheskinonscannedhistopathologicaltissuesections |