Automated cutaneous squamous cell carcinoma grading using deep learning with transfer learning

• DOI: 10.47162/rjme.65.2.10
• Publication Date: 2024-08-07T08:57:22Z
• Authors: Alexandra Buruiană; Mircea-Sebastian ...; ; Bogdan Pop; ; Bogdan-Alexandru ...; ; Carmen Georgiu; ; Doiniţa Crişan; ; Maria Crişan
• Keywords: Grading (engineering); Transfer of learning

Introduction: Histological grading of cutaneous squamous cell carcinoma (cSCC) is crucial for prognosis and treatment decisions, but manual subjective time-consuming. Aim: This study aimed to develop validate a deep learning (DL)-based model automated cSCC grading, potentially improving diagnostic accuracy (ACC) efficiency. Materials Methods: Three neural networks (DNNs) with different architectures (AlexNet, GoogLeNet, ResNet-18) were trained using transfer on dataset 300 histopathological images cSCC. The models evaluated their ACC, sensitivity (SN), specificity (SP), area under the curve (AUC). Clinical validation was performed 60 images, comparing DNNs’ predictions those panel pathologists. Results: achieved high performance metrics (ACC>85%, SN>85%, SP>92%, AUC>97%) demonstrating potential objective efficient grading. agreement between DNNs pathologists, as well among network architectures, further supports reliability ACC DL models. top-performing are publicly available, facilitating research clinical implementation. Conclusions: highlights promising role in enhancing diagnosis, ultimately patient care.