Clinically guided trainable soft attention for early detection of oral cancer
Document Type
Conference Item
Publication Date
1-1-2021
Abstract
Oral cancer disproportionately affects low- and middle-income countries, where a lack of access to appropriate medical care contributes towards late disease presentation. Using artificial intelligence to facilitate the automated identification of high-risk oral lesions can improve patient survival rates. With image classification using oral cavity images and other forms of medical images, the information to be classified can often be extremely localized. To address this problem, we propose the use of convolutional neural networks with trainable soft attention. Further to this, we incorporate the use of localization loss to penalize the difference between attention maps and clinically annotated mask. This effectively allows clinicians to help guide soft attention. Improvements to the baseline were made, with an accuracy of 0.8333 and a ROC AUC of 0.8632, which equates to increases of 0.0245 and 0.0394, respectively. This accuracy corresponds to a sensitivity of 0.8469 and a specificity of 0.8208. Perhaps of more importance, is a model that demonstrates better capability at paying attention to the lesions in its decision making. Furthermore, visualizing resulting attention maps can help to strengthen clinical confidence in AI decision making. © 2021, Springer Nature Switzerland AG.
Keywords
Attention, Deep learning, Oral cancer, Oral potentially malignant disorders
Divisions
Dentistry
Funders
Medical Research Council [Grant No: MR/S013865/1]
Publication Title
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume
13052
Event Title
19th International Conference on Computer Analysis of Images and Patterns, CAIP 2021
Event Location
Virtual, Online
Event Dates
28 - 30 September 2021
Event Type
conference