Medical image data acquired from ultrasound, X-rays (CT), MR and other types of imaging
modalities is routinely used by doctors in detecting, diagnosing and planning treatment for
myriad diseases. The problem of missing data is ubiquitous in medical imaging and leads
to the loss in utility of the images, and an accompanying loss in the accuracy of detection,
diagnosis and treatment planning for a disease. As an example, in the management of
renal masses, there is significant loss of CECT phase images (greater than 40%) and
this leads to less accurate diagnosis of malignant masses and adversely effects surgical
planning in cases where surgical intervention is necessary.
In this MHI grant we will address this problem by developing a statistical deep-learning
based technique for imputing missing phase images in a sequence of renal CECT images.
This technique will provide the best guess for the missing phase and also quantify the
confidence in this guess. Its accuracy will be throughly tested and its efficacy in improving
the management of renal masses will be quantified. We note that the concept is general
and not tied to a specific image feature or type, and therefore can be applied to imputing
medical images of any type.
The proposed goal will be achieved by three labs from the Viterbi and Keck schools with
complimentary expertise in computational science and data and physics-driven modeling,
abdominal imaging and radiomics, and urology. Future plans for this research involve
a larger clinical study sponsored by the NIH followed by the translation of the proposed
technology to clinical practice.
