Automated classification of oral potentially malignant disorders and oral squamous cell carcinoma using a convolutional neural network framework: a cross-sectional study

IF 7 Q1 HEALTH CARE SCIENCES & SERVICES

Lancet Regional Health-Americas Pub Date : 2025-05-29 DOI:10.1016/j.lana.2025.101138

Cristina Saldivia-Siracusa , Eduardo Santos Carlos de Souza , Arnaldo Vitor Barros da Silva , Anna Luíza Damaceno Araújo , Caíque Mariano Pedroso , Tarcília Aparecida da Silva , Maria Sissa Pereira Sant'Ana , Felipe Paiva Fonseca , Hélder Antônio Rebelo Pontes , Marcos G. Quiles , Marcio Ajudarte Lopes , Pablo Agustin Vargas , Syed Ali Khurram , Alexander T. Pearson , Mark W. Lingen , Luiz Paulo Kowalski , Keith D. Hunter , André Carlos Ponce de Leon Ferreira de Carvalho , Alan Roger Santos-Silva

{"title":"Automated classification of oral potentially malignant disorders and oral squamous cell carcinoma using a convolutional neural network framework: a cross-sectional study","authors":"Cristina Saldivia-Siracusa , Eduardo Santos Carlos de Souza , Arnaldo Vitor Barros da Silva , Anna Luíza Damaceno Araújo , Caíque Mariano Pedroso , Tarcília Aparecida da Silva , Maria Sissa Pereira Sant'Ana , Felipe Paiva Fonseca , Hélder Antônio Rebelo Pontes , Marcos G. Quiles , Marcio Ajudarte Lopes , Pablo Agustin Vargas , Syed Ali Khurram , Alexander T. Pearson , Mark W. Lingen , Luiz Paulo Kowalski , Keith D. Hunter , André Carlos Ponce de Leon Ferreira de Carvalho , Alan Roger Santos-Silva","doi":"10.1016/j.lana.2025.101138","DOIUrl":null,"url":null,"abstract":"<div><h3>Background</h3><div>Artificial Intelligence (AI) models hold promise as useful tools in healthcare practice. We aimed to develop and assess AI models for automatic classification of oral potentially malignant disorders (OPMD) and oral squamous cell carcinoma (OSCC) clinical images through a Deep Learning (DL) approach, and to explore explainability using Gradient-weighted Class Activation Mapping (Grad-CAM).</div></div><div><h3>Methods</h3><div>This study assessed a dataset of 778 clinical images of OPMD and OSCC, divided into training, model optimization, and internal testing subsets with an 8:1:1 proportion. Transfer learning strategies were applied to pre-train 8 convolutional neural networks (CNN). Performance was evaluated by mean accuracy, precision, recall, specificity, F1-score and area under the receiver operating characteristic (AUROC) values. Grad-CAM qualitative appraisal was performed to assess explainability.</div></div><div><h3>Findings</h3><div>ConvNeXt and MobileNet CNNs showed the best performance. Transfer learning strategies enhanced performance for both algorithms, and the greatest model achieved mean accuracy, precision, recall, F1-score and AUROC of 0.799, 0.837, 0.756, 0.794 and 0.863 during internal testing, respectively. MobileNet displayed the lowest computational cost. Grad-CAM analysis demonstrated discrepancies between the best-performing model and the highest explainability model.</div></div><div><h3>Interpretation</h3><div>ConvNeXt and MobileNet DL models accurately distinguished OSCC from OPMD in clinical photographs taken with different types of image-capture devices. Grad-CAM proved to be an outstanding tool to improve performance interpretation. Obtained results suggest that the adoption of DL models in healthcare could aid in diagnostic assistance and decision-making during clinical practice.</div></div><div><h3>Funding</h3><div>This work was supported by <span>FAPESP</span> (2022/13069-8, 2022/07276-0, 2021/14585-7 and 2024/20694-1), <span>CAPES</span>, <span>CNPq</span> (307604/2023-3) and <span>FAPEMIG</span>.</div></div>","PeriodicalId":29783,"journal":{"name":"Lancet Regional Health-Americas","volume":"47 ","pages":"Article 101138"},"PeriodicalIF":7.0000,"publicationDate":"2025-05-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Lancet Regional Health-Americas","FirstCategoryId":"1085","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2667193X25001486","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"HEALTH CARE SCIENCES & SERVICES","Score":null,"Total":0}

引用次数: 0

Abstract

Background

Artificial Intelligence (AI) models hold promise as useful tools in healthcare practice. We aimed to develop and assess AI models for automatic classification of oral potentially malignant disorders (OPMD) and oral squamous cell carcinoma (OSCC) clinical images through a Deep Learning (DL) approach, and to explore explainability using Gradient-weighted Class Activation Mapping (Grad-CAM).

Methods

This study assessed a dataset of 778 clinical images of OPMD and OSCC, divided into training, model optimization, and internal testing subsets with an 8:1:1 proportion. Transfer learning strategies were applied to pre-train 8 convolutional neural networks (CNN). Performance was evaluated by mean accuracy, precision, recall, specificity, F1-score and area under the receiver operating characteristic (AUROC) values. Grad-CAM qualitative appraisal was performed to assess explainability.

Findings

ConvNeXt and MobileNet CNNs showed the best performance. Transfer learning strategies enhanced performance for both algorithms, and the greatest model achieved mean accuracy, precision, recall, F1-score and AUROC of 0.799, 0.837, 0.756, 0.794 and 0.863 during internal testing, respectively. MobileNet displayed the lowest computational cost. Grad-CAM analysis demonstrated discrepancies between the best-performing model and the highest explainability model.

Interpretation

ConvNeXt and MobileNet DL models accurately distinguished OSCC from OPMD in clinical photographs taken with different types of image-capture devices. Grad-CAM proved to be an outstanding tool to improve performance interpretation. Obtained results suggest that the adoption of DL models in healthcare could aid in diagnostic assistance and decision-making during clinical practice.

Funding

This work was supported by FAPESP (2022/13069-8, 2022/07276-0, 2021/14585-7 and 2024/20694-1), CAPES, CNPq (307604/2023-3) and FAPEMIG.

查看原文本刊更多论文

使用卷积神经网络框架自动分类口腔潜在恶性疾病和口腔鳞状细胞癌：一项横断面研究

人工智能（AI）模型有望成为医疗保健实践中的有用工具。我们旨在通过深度学习（DL）方法开发和评估用于口腔潜在恶性疾病（OPMD）和口腔鳞状细胞癌（OSCC）临床图像自动分类的AI模型，并使用梯度加权类激活映射（Grad-CAM）探索可解释性。方法选取778张OPMD和OSCC临床影像数据集，按8:1:1的比例分为训练子集、模型优化子集和内部测试子集。应用迁移学习策略对8个卷积神经网络（CNN）进行预训练。通过平均准确度、精密度、召回率、特异性、f1评分和受试者操作特征（AUROC）值下的面积来评价疗效。采用Grad-CAM定性评价来评估可解释性。发现convnext和MobileNet cnn表现出最好的性能。迁移学习策略提高了两种算法的性能，在内部测试中，最大模型的平均准确率、精密度、召回率、f1得分和AUROC分别为0.799、0.837、0.756、0.794和0.863。MobileNet显示的计算成本最低。Grad-CAM分析显示了最佳表现模型和最高可解释性模型之间的差异。解释：convnext和MobileNet DL模型在使用不同类型的图像捕获设备拍摄的临床照片中准确区分了OSCC和OPMD。事实证明，Grad-CAM是一种改进性能解释的出色工具。所得结果表明，在医疗保健中采用深度学习模型有助于临床实践中的诊断辅助和决策。本研究得到了FAPESP（2022/13069- 8,2022 /07276- 0,2021 /14585-7和2024/20694-1）、CAPES、CNPq（307604/2023-3）和FAPEMIG的支持。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Lancet Regional Health-Americas Multiple-

CiteScore

8.00

自引率

0.00%

发文量

期刊介绍： The Lancet Regional Health – Americas, an open-access journal, contributes to The Lancet's global initiative by focusing on health-care quality and access in the Americas. It aims to advance clinical practice and health policy in the region, promoting better health outcomes. The journal publishes high-quality original research advocating change or shedding light on clinical practice and health policy. It welcomes submissions on various regional health topics, including infectious diseases, non-communicable diseases, child and adolescent health, maternal and reproductive health, emergency care, health policy, and health equity.