Abstract
Recently, Raman Spectroscopy (RS) was demonstrated to be a non-destructive way of cancer diagnosis, due to the uniqueness of RS measurements in revealing molecular biochemical changes between cancerous vs. normal tissues and cells. In order to design computational approaches for cancer detection, the quality and quantity of tissue samples for RS are important for accurate prediction. In reality, however, obtaining skin cancer samples is difficult and expensive due to privacy and other constraints. With a small number of samples, the training of the classifier is difficult, and often results in overfitting. Therefore, it is important to have more samples to better train classifiers for accurate cancer tissue classification. To overcome these limitations, this paper presents a novel generative adversarial network based skin cancer tissue classification framework. Specifically, we design a data augmentation module that employs a Generative Adversarial Network (GAN) to generate synthetic RS data resembling the training data classes. The original tissue samples and the generated data are concatenated to train classification modules. Experiments on real-world RS data demonstrate that (1) data augmentation can help improve skin cancer tissue classification accuracy, and (2) generative adversarial network can be used to generate reliable synthetic Raman spectroscopic data.
Highlights
Despite of the promising properties, Raman Spectroscopy (RS) suffers from weak signals due to inherent noise
Deep learning for RS cancer tissue classification By leveraging sample augmentation, we propose a deep learning based framework and compare its performance using a variety of deep learning models, including convolutional neural network (CNN) and Long Short Term Memory Networks (LSTM), for Raman spectroscopy cancer tissue classification
We have following observations: (1) Most deep network models, such as CNN and LSTM, outperform traditional machine learning models, like logistic regression (LR) and support vector machines (SVM). This demonstrates that deep learning methods, even with limited training samples, can learn better hidden representations of samples. This is because that deep learning can better leverage correlation in the RS signals to learn patterns, whereas LR and SVM take raw RS spectra as features, where feature correlation will deteriorate the classifier performance
Summary
Despite of the promising properties, RS suffers from weak signals due to inherent noise. The study systematically varied the ablation level and examined its impact, which showed that Raman spectral features from normal and cancerous tissue did not significantly correlate with ablation treatment level (in this study, we combine both laser treated and untreated samples to maximize number of available samples). Another challenge in RS based skin cancer detection is that the data are often ill-posed, because Raman spectrum has a background resulted from skin fluorescence, both the spectra and correlation can be introduced with variance.
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
