Article
Prediction of melanoma Breslow thickness using deep transfer learning algorithms
Author/s | Hernández Rodríguez, Juan Carlos
Durán López, Lourdes Domínguez Morales, Juan Pedro Ortiz-Álvarez, Juan Conejo-Mir Sánchez, Julián Pereyra-Rodríguez, José-Juan |
Department | Universidad de Sevilla. Departamento de Medicina Universidad de Sevilla. Departamento de Arquitectura y Tecnología de Computadores |
Publication Date | 2023-07 |
Deposit Date | 2024-02-26 |
Published in |
|
Abstract | Background:
The distinction between in situ melanoma (MIS) and invasive melanoma is challenging even for expert dermatologists. The use of pretrained convolutional neural networks (CNNs) as ancillary decision systems ... Background: The distinction between in situ melanoma (MIS) and invasive melanoma is challenging even for expert dermatologists. The use of pretrained convolutional neural networks (CNNs) as ancillary decision systems needs further research. Aim To develop, validate and compare three deep transfer learning (DTL) algorithms to predict MIS vs. invasive melanoma and melanoma with a Breslow thickness (BT) of < 0.8 mm vs. ≥ 0.8 mm. Methods: A dataset of 1315 dermoscopic images of histopathologically confirmed melanomas was created from Virgen del Rocio University Hospital and open repositories of the International Skin Imaging Collaboration archive and Polesie S et al. (Dermatol Pract Concept 2021; 11:e2021079). The images were labelled as MIS or invasive melanoma and < 0.8 mm or ≥ 0.8 mm of BT. We conducted three trainings, and overall means for receiver operating characteristic (ROC) curves, sensitivity, specificity, positive and negative predictive value, and balanced diagnostic accuracy outcomes were evaluated on the test set with ResNetV2, EfficientNetB6 and InceptionV3. The results of 10 dermatologists were compared with the algorithms. Grad-CAM gradient maps were generated, highlighting relevant areas considered by the CNNs within the images. Results: EfficientNetB6 achieved the highest diagnostic accuracy for the comparison between MIS vs. invasive melanoma (61%) and BT < 0.8 mm vs. ≥ 0.8 mm (75%). For the BT comparison, ResNetV2 with an area under the ROC curve of 0.76 and InceptionV3 with an area under the ROC curve of 0.75, outperformed the results obtained by the dermatologist group with an area under the ROC curve of 0.70. Conclusion: EfficientNetB6 recorded the best prediction results, outperforming the dermatologists for the comparison of 0.8 mm of BT. DTL could be an ancillary aid to support dermatologists’ decisions in the near future. |
Citation | Hernández Rodríguez, J.C., Durán López, L., Domínguez Morales, J.P., Ortiz-Álvarez, J., Conejo-Mir Sánchez, J. y Pereyra-Rodríguez, J. (2023). Prediction of melanoma Breslow thickness using deep transfer learning algorithms. Clinical and Experimental Dermatology, 48 (7), 752-758. https://doi.org/10.1093/ced/llad107. |
Files | Size | Format | View | Description |
---|---|---|---|---|
Prediction of melanoma Breslow ... | 2.740Mb | [PDF] | This document is not available in full text until 2024-07 . For more information contact idus@us.es. | |
This item appears in the following collection(s)
This document is protected by intellectual and industrial property rights. Without prejudice to existing legal exemptions, its reproduction, distribution, public communication or transformation is prohibited without the authorization of the rights holder, unless otherwise indicated.