Role of deep-learning fusion models of optical and ultrasound imaging in the diagnosis of cancer and assessment of cancer treatment response

Abstract: The past decade has seen remarkable progress in deep-learning neural networks (DNNs) in medicine.  In medical imaging, unlike traditional machine learning methods that require manually extracting features from input images, DNNs learn the complex image features needed for the diagnosis of diseases, such as cancer. However, for any new imaging modality, such as optical imaging, data are often limited to robustly train DNNs.  Fusion models that combine DNNs of conventional imaging modalities with new imaging modalities can significantly improve the diagnostic performance of DNNs trained from imaging data of each modality alone. In this talk, I will review our recent progress in developing fusion models of optical imaging and ultrasound imaging for the diagnosis of breast cancer and ovarian cancer, and for the assessment of the treatment response of rectal cancer and breast cancer.  Because physicians are familiar with conventional imaging modalities, they can quickly adopt fusion models as intelligent decision support systems.

Role of deep-learning fusion models of optical and ultrasound imaging in the diagnosis of cancer and assessment of cancer treatment response

Abstract: The past decade has seen remarkable progress in deep-learning neural networks (DNNs) in medicine.  In medical imaging, unlike traditional machine learning methods that require manually extracting features from input images, DNNs learn the complex image features needed for the diagnosis of diseases, such as cancer. However, for any new imaging modality, such as optical imaging, data are often limited to robustly train DNNs.  Fusion models that combine DNNs of conventional imaging modalities with new imaging modalities can significantly improve the diagnostic performance of DNNs trained from imaging data of each modality alone. In this talk, I will review our recent progress in developing fusion models of optical imaging and ultrasound imaging for the diagnosis of breast cancer and ovarian cancer, and for the assessment of the treatment response of rectal cancer and breast cancer.  Because physicians are familiar with conventional imaging modalities, they can quickly adopt fusion models as intelligent decision support systems.