Artificial Intelligence for Image Analysis in Oral Squamous Cell Carcinoma: A Review

Pereira-Prado, Vanesa; Martins-Silveira, Felipe; Sicco, Estafanía; Hochmann, Jimena; Isiordia-Espinoza, Mario; González, Rogelio; Pandiar, Deepak; Bologna-Molina, Ronell

doi:10.3390/diagnostics13142416

Cited by 4 publications

(1 citation statement)

References 56 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Briefly, DL, a part of the wide field of AI, may be divided into two DL-based unsupervised feature learning and handcrafted approach [ 11 ]. With the application of convoluted neural network (CNN), in the former, segmentation of images, object detection, and classification have been attempted for OSCC [ 12 ]. Various CNNs are available for the task such as InceptionV3, U-Net, ResNet-50, and AlexNet, to name a few [ 11 ].…”

Section: Introductionmentioning

confidence: 99%

Application of InceptionV3, SqueezeNet, and VGG16 Convoluted Neural Networks in the Image Classification of Oral Squamous Cell Carcinoma: A Cross-Sectional Study

Pandiar,

Choudhari,

Poothakulath Krishnan

2023

Cureus

Self Cite

View full text Add to dashboard Cite

Background Artificial intelligence (AI) is a rapidly emerging field in medicine and has applications in diagnostics, therapeutics, and prognostication in various malignancies. The present study was conducted to analyze and compare the accuracy of three deep learning neural networks for oral squamous cell carcinoma (OSCC) images. Materials and methods Three hundred and twenty-five cases of OSCC were included and graded histologically by two grading systems. The images were then analyzed using the Orange data mining tool. Three neural networks, viz., InceptionV3, SqueezeNet, and VGG16, were used for further analysis and classification. Positive predictive value, negative predictive value, specificity, sensitivity, area under curve (AUC), and accuracy were estimated for each neural network. Results Histological grading by Bryne's yielded significantly stronger inter-observer agreement. The highest accuracy was found for the classification of poorly differentiated squamous cell carcinoma images irrespective of the network used. Other values were variegated. Conclusion AI could serve as an adjunct for improvement in theragnostics. Further research is required to achieve the modification of mining tools for greater predictive values, sensitivity, specificity, AUC, accuracy, and security. Bryne's grading system is warranted for the better application of AI in OSCC image analytics.

show abstract

Section: Introductionmentioning

confidence: 99%