Accurate delineation of tumors and sensitive structures is important for many medical applications. One example is treatment planning in cancer radiotherapy. Although many auto-segmentation algorithms have been developed and implemented in clinical practice, none of them are satisfactory and manual contouring is often required. Such failure is mainly because conventional segmentation methods are purely based on local information in the images. We have been investigating deep learning based strategies to solve this problem.
One of our projects is to use a deep-learning model for brain metastasis tumor segmentation. Another project is about cervical tumor segmentation in PET images using CNN. In PET, cervical tumors are often connected to bladders. Traditional algorithms such as region growing and thresholding based algorithms often misclassify bladder as tumor. We are also building a labeled patient image database, AnatomyNet. Such database will provide a large train dataset to establish a sophisticated CNN model for auto-segmentation.
- Liu, Y., Stojadinovic, S., Hrycushko, B., Wardak, Z., Lu, W., Yan, Y., ... & Gu, X. (2016). Automatic metastatic brain tumor segmentation for stereotactic radiosurgery applications. Physics in medicine and biology, 61(24), 8440. (journal)(pdf)
- Liu, Y., Stojadinovic, S., Hrycushko, B., Wardak, Z., Lau, S., Lu, W., ... & Nedzi, L. (2017). A deep convolutional neural network-based automatic delineation strategy for multiple brain metastases stereotactic radiosurgery. PloS one, 12(10), e0185844. (journal)(pdf)
- Rozario, T., Long, T., Chen, M., Lu, W., & Jiang, S. (2017). Towards automated patient data cleaning using deep learning: A feasibility study on the standardization of organ labeling. arXiv preprint arXiv:1801.00096. (arXiv)