Project involves the extension of deep learning/AI methods, which we already have in the lab, to the assessment of various disease types as presented on medical images. For example, the classification of breast cancer subtypes and prognosis from mammography, ultrasound, and/or MRI, the quantitative assessment of disease on low-dose CT, and the monitoring of response to therapy (such as in traumatic brain injury).

The aim of the research is to demonstrate the ability of deep learning (convolutional neural networks) in the assessment of disease as compared to human interpretations and/or conventional computer-aided diagnostic methods.

My radiology computer vision/AI lab has been investigating methods for the quantitative analysis of medical images for the past few decades. This is now termed 'radiomics', which involves the conversion of medical images to mineable data. We have developed and clinically translated various methods in breast cancer imaging and assessment, and have related these radiomic features of breast tumors to genomics in collaboration with the NCI TCGA. More recently we have implemented various deep learning AI algorithms using convolutional neural networks on breast, skeletal, and thoracic images that complement the traditional CAD algorithms. The student will work with my lab members and clinical collaborators in (1) identifying their specific project, i.e., clinical question, (2) collecting the database, (3) running existing computer image analysis programs to extract radiomic features, and (4) evaluate these deep learning methods relevant to the clinical question. Student will attend lab project meetings and have the opportunity to author abstracts to national meetings.

All software available in the Giger lab

RSNA, AUR, ASCO