Method of histopathological diagnosis of mammary nodules through deep learning algorithm
ABSTRACT Introduction: Artificial intelligence systems are promising health care technologies, mainly in medical subareas such as pathology, and can be used as support methods for the histological diagnosis of mammary nodules. Objective: This study describes the method and results of the develop...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Sociedade Brasileira de Patologia Clínica
2020-03-01
|
| Series: | Jornal Brasileiro de Patologia e Medicina Laboratorial |
| Subjects: | |
| Online Access: | http://www.scielo.br/scielo.php?script=sci_arttext&pid=S1676-24442019000600620&tlng=pt |
| _version_ | 1798025624056496128 |
|---|---|
| author | Victor Antônio Kuiava Eliseu Luiz Kuiava Rubens Rodriguez Adriana Eli Beck João Pedro M. Rodriguez Eduardo O. Chielle |
| author_facet | Victor Antônio Kuiava Eliseu Luiz Kuiava Rubens Rodriguez Adriana Eli Beck João Pedro M. Rodriguez Eduardo O. Chielle |
| author_sort | Victor Antônio Kuiava |
| collection | DOAJ |
| description | ABSTRACT Introduction: Artificial intelligence systems are promising health care technologies, mainly in medical subareas such as pathology, and can be used as support methods for the histological diagnosis of mammary nodules. Objective: This study describes the method and results of the development of artificial intelligence software for the histopathological analysis of mammary nodules. Methods: The software was developed by using two neural networks - Inception and MobileNet. The database used for learning the conditions analyzed (histologically normal breast, fibroadenoma, fibrocystic changes, in situ ductal carcinoma, invasive carcinoma of no special type and invasive lobular carcinoma) was obtained after authorization of the Path Presenter site with 5,298 images. The 2,740 images used for the validation of the system were obtained from the Pathology Institute of Passo Fundo. Results: The present software had sensitivity of 80.5% [95% confidence interval (CI), 71.9%-89.1%] and specificity of 96.1% (95% CI, 94.3%-97.8%) for MobileNet and sensitivity of 73.8% (95% CI, 52.6%-115%) and specificity of 94.7% (CI 95%, 91.7%-97.7%) for Inception. For the differentiation of malignant conditions, it obtained a maximum sensitivity of 78.7% and specificity of 95.8%; for differentiation of benign conditions, the maximum sensitivity was 82.6% and the specificity was 97.4%. Conclusion: The present software presented promising results in the histopathological analysis of mammary nodules. It reinforced the idea that in the future the presence of diagnostic support systems in breast pathologies may play a crucial role in health care. |
| first_indexed | 2024-04-11T18:21:51Z |
| format | Article |
| id | doaj.art-2c4995671ed14f3b9b44a29214090be8 |
| institution | Directory Open Access Journal |
| issn | 1678-4774 |
| language | English |
| last_indexed | 2024-04-11T18:21:51Z |
| publishDate | 2020-03-01 |
| publisher | Sociedade Brasileira de Patologia Clínica |
| record_format | Article |
| series | Jornal Brasileiro de Patologia e Medicina Laboratorial |
| spelling | doaj.art-2c4995671ed14f3b9b44a29214090be82022-12-22T04:09:45ZengSociedade Brasileira de Patologia ClínicaJornal Brasileiro de Patologia e Medicina Laboratorial1678-47742020-03-0155662063210.5935/1676-2444.20190055Method of histopathological diagnosis of mammary nodules through deep learning algorithmVictor Antônio KuiavaEliseu Luiz KuiavaRubens RodriguezAdriana Eli BeckJoão Pedro M. RodriguezEduardo O. Chiellehttps://orcid.org/0000-0003-3566-1258ABSTRACT Introduction: Artificial intelligence systems are promising health care technologies, mainly in medical subareas such as pathology, and can be used as support methods for the histological diagnosis of mammary nodules. Objective: This study describes the method and results of the development of artificial intelligence software for the histopathological analysis of mammary nodules. Methods: The software was developed by using two neural networks - Inception and MobileNet. The database used for learning the conditions analyzed (histologically normal breast, fibroadenoma, fibrocystic changes, in situ ductal carcinoma, invasive carcinoma of no special type and invasive lobular carcinoma) was obtained after authorization of the Path Presenter site with 5,298 images. The 2,740 images used for the validation of the system were obtained from the Pathology Institute of Passo Fundo. Results: The present software had sensitivity of 80.5% [95% confidence interval (CI), 71.9%-89.1%] and specificity of 96.1% (95% CI, 94.3%-97.8%) for MobileNet and sensitivity of 73.8% (95% CI, 52.6%-115%) and specificity of 94.7% (CI 95%, 91.7%-97.7%) for Inception. For the differentiation of malignant conditions, it obtained a maximum sensitivity of 78.7% and specificity of 95.8%; for differentiation of benign conditions, the maximum sensitivity was 82.6% and the specificity was 97.4%. Conclusion: The present software presented promising results in the histopathological analysis of mammary nodules. It reinforced the idea that in the future the presence of diagnostic support systems in breast pathologies may play a crucial role in health care.http://www.scielo.br/scielo.php?script=sci_arttext&pid=S1676-24442019000600620&tlng=ptbreast cancerartificial intelligencediagnosis |
| spellingShingle | Victor Antônio Kuiava Eliseu Luiz Kuiava Rubens Rodriguez Adriana Eli Beck João Pedro M. Rodriguez Eduardo O. Chielle Method of histopathological diagnosis of mammary nodules through deep learning algorithm Jornal Brasileiro de Patologia e Medicina Laboratorial breast cancer artificial intelligence diagnosis |
| title | Method of histopathological diagnosis of mammary nodules through deep learning algorithm |
| title_full | Method of histopathological diagnosis of mammary nodules through deep learning algorithm |
| title_fullStr | Method of histopathological diagnosis of mammary nodules through deep learning algorithm |
| title_full_unstemmed | Method of histopathological diagnosis of mammary nodules through deep learning algorithm |
| title_short | Method of histopathological diagnosis of mammary nodules through deep learning algorithm |
| title_sort | method of histopathological diagnosis of mammary nodules through deep learning algorithm |
| topic | breast cancer artificial intelligence diagnosis |
| url | http://www.scielo.br/scielo.php?script=sci_arttext&pid=S1676-24442019000600620&tlng=pt |
| work_keys_str_mv | AT victorantoniokuiava methodofhistopathologicaldiagnosisofmammarynodulesthroughdeeplearningalgorithm AT eliseuluizkuiava methodofhistopathologicaldiagnosisofmammarynodulesthroughdeeplearningalgorithm AT rubensrodriguez methodofhistopathologicaldiagnosisofmammarynodulesthroughdeeplearningalgorithm AT adrianaelibeck methodofhistopathologicaldiagnosisofmammarynodulesthroughdeeplearningalgorithm AT joaopedromrodriguez methodofhistopathologicaldiagnosisofmammarynodulesthroughdeeplearningalgorithm AT eduardoochielle methodofhistopathologicaldiagnosisofmammarynodulesthroughdeeplearningalgorithm |