Mapping Chemo-Brain and Predicting Long-Term Cognitive Function of Breast Cancer Survivors Using Multiple Mri Features Machine Learning

Recent advance in primary treatment of breast cancer marks the importance of the post-treatment care. Cancer-related trauma (after operation and chemotherapy) has been widely reported by breast cancer survivors. Given that the cancer traumatic experience contains multiple and chronic stressors during diagnosis and treatment courses, i.e. post-traumatic stress disorder (PTSD), it differs from other kind of traumatic event. Previous research has demonstrated smaller hippocampus, amygdala, and prefrontal cortex in individuals with post-traumatic stress disorder, and smaller frontal, temporal lobe as well as decreased cognitive function in individuals after chemotherapy. Although, our preliminary results showed that altered functional activity in the dorsal attention network and decline of cerebral white matter in midterm post-chemotherapy patients and potentially related to clinical cognitive assessments, there remains a paucity of advanced neuroimaging studies in chemotherapy survivors and cognitive function prediction. The aim of this prospective neuroimaging study is to assess the longitudinal changes of chemo-brain structure and function following diagnosis of breast cancer, and to evualuate if the brain magnetic resonance imaging (MRI) acquired at pre-treatment baseline or post-chemotherapy can accurately predict breast cancer-related cognitive outcome at long-term follow-up using machine learning. Thirty-five patients with breast cancer before chemotherapy (BC), 35 patients after chemotherapy (AC), 35 patients after hormone therapy (AH) and 35 healthy controls (HC) will be recruited in this study. Participants in BC (N=30) and HC groups (N=30) will be follow up at 6-24 months. MRI scans and psychometrics will be used to assess all participants, and the MRI will include generalized q-sampling MRI (GQI) and functional connectivity. We will evaluate patients with breast cancer prior to any treatment, post-chemotherapy and 6-24 months later. Cognitive testing scores will be normalized based on data obtained from the healthy female controls and then will be used to categorize patients as impaired or not based on longitudinal changes using machine learning and support vector machine. Previous studies on structural brain changes in cancer-related trauma and functional brain changes in chemotherapy were based on cross-sectional design and demonstrated controversial results. The strengths of our study include: 1. longitudinal study design to survey brain structure and function alteration after diagnosis, chemotherapy, and follow-up; 2. analysis of the relation of brain structure and function changes to breast cancer-related traumatic events, chemotherapy and psychometrics; 3. advanced MR imaging (GQI, functional connectivity) and analysis techniques (machine learning, support vector machine) will be used. 4. machine learning will be perform to predict the cognitive coutcome of the survivors after chemotherapy. Hopefully, the results can provide the evidence of brain structural and functional changes in women with breast cancer, and predict future cognitive impairment associated with breast cancer after chemotherapy. This information can highlight the importance of the breast cancer-related PTSD and chemotherapy, and inform treatment decision making by identifying patients at highest risk for long-term cognitive impairment.

Project ID:PB10901-1019
External Project ID:MOST107-2221-E182-054-MY3