Classification of quantitative light-induced fluorescence images using convolutional neural network

Images are an important data source for diagnosis of oral diseases. The manual classification of images may lead to suboptimal treatment procedures due to subjective errors. In this paper an image classification algorithm based on Deep Learning framework is applied to Quantitative Light-induced Fluorescence (QLF) images [4]. The Convolutional Neural Network [3] (CNN) outperforms other state of the art shallow classification models in predicting labels derived from three different dental plaque assessment scores. Such result is possible because our model directly learns invariant feature representations from raw pixel intensity values without any hand-crafted feature engineering. The model benefits from multi-channel representation of the images resulting in improved performance when, besides the Red colour channel, additional Green and Blue colour channels are used. Previous studies on this topic either focused on only single plaque scoring system without providing detailed analysis of results [1] or used a smaller dataset of non-QLF images and a shallow network architecture [2] to address the problem. We expect that Deep Learning of QLF-images can help dental practitioners to perform efficient plaque assessments and contribute to the improvement of patients’ oral health. An extended version of the manuscript with detailed description of the experimental setup and the obtained results can be found at http://arxiv.org/abs/1705.09193 or http://learning-machines.com/.