Poor prognosis with short median overall survival of 11 months of glioblastoma patients (GBM, WHO grade IV) has been a prominent problem in brain cancer field. One of the major challenges is the lack of accurate tool for identification of GBM status post chemo-radiation (SOC), which is often ambiguous. The limited reliability of MRIs in determination of GBM progression has been a big issue for physicians to decide when to change treatment. Although advanced brain tumor imaging MRI technologies (ABTI MRI) helps, their cost and availability make them not accessible to most patients, especially in non-academic centers. There are also data, such as information electronic health records and molec...

Read More

Poor prognosis with short median overall survival of 11 months of glioblastoma patients (GBM, WHO grade IV) has been a prominent problem in brain cancer field. One of the major challenges is the lack of accurate tool for identification of GBM status post chemo-radiation (SOC), which is often ambiguous. The limited reliability of MRIs in determination of GBM progression has been a big issue for physicians to decide when to change treatment. Although advanced brain tumor imaging MRI technologies (ABTI MRI) helps, their cost and availability make them not accessible to most patients, especially in non-academic centers. There are also data, such as information electronic health records and molecular biomarkers, which are not jointly utilized in the determination of GBM patient care. Realizing the challenge and the technological gap, we are coming up with Artificial Intelligence (AI) technologies to connect the dots by fusing complementary information in order to build a stronger and better model for GBM responses to SOC. Highlighting on co-teaching algorithms to learn from ABTI MRI and multi-modality learning algorithm to identify hybrid biomarkers as well as obtaining tumor volume with no or minimal human supervision, we are piloting novel solutions to push the fronts of GBM research and improving patient care. As a group of highly dedicated and experienced physicians and informaticians, we plan to develop integrated AI solutions with human-in-the-loop. Built on the clinical knowledge and biological findings, our novel machine learning algorithm is different from off-the-shelf tools and we will consider the inherit connections between healthcare data and to address the root of the challenges and designed to learn from fragmented information. If successful, the project will make a profound improvement on GBM patient care, biomarker research and clinical trials. It will open doors to apply AI to the practical use of other brain tumor research.

Read Less