Longitudinal Analysis of Mammographic Breast Lesions and their Microenvironments
Q. How did you become involved in cancer research?
A. I started collaborating with Dr. Kevin Mills on lymphoma/leukemia at The Jackson Lab, and started working on computational techniques to study mammographic breast lesions at the same time. I also have a personal connection to cancer. My cousin Elizabeth was diagnosed with bone cancer at age 8. She had to have her whole leg amputated and undergo aggressive chemotherapy. She is fine today. My brother, Jean-Francois, is currently battling a rare form of Hodgkin’s lymphoma. He was diagnosed at age 39 and had been in remission, but the cancer came back and he started a new, more aggressive chemotherapy treatment in September 2012. I'm pleased to report that he is in remission today.
Q. Please describe your current work:
A. We are trying to obtain a better understanding of the structure and dynamics of breast tissue lesions, and specifically how malignant tumors appear and lead to cancer. One of our key recent findings shows that there could be a specific breast tissue physical state that is prone to tumor development. If validated, we could be able to determine specific regions of the breast that are prone to the development of tumors. This would lead not only to earlier breast cancer detection, but perhaps even “pre-detection”.
Q. What’s the most exciting thing on the horizon?
A. For many in our field, the most exciting thing on the horizon is to be able to detect existing breast tumors before the radiologists can detect them visually. We think that it would be much better to be able to detect tumor-prone regions in the breast even before a tumor appears.
Q. What would you say to someone recently given a cancer diagnosis?
A. I would say that the treatment options and strategies that you receive today were theoretical and experimental 5-10+ years ago. Your chances of survival are that much better than in the past. The other thing I would say is: make sure you have the moral support you need and surround yourself with loved ones when you fight the fight. The psychological aspect of the fight against cancer is key.
Maine Cancer Foundation Funded Research Proposal:
Over the course of a lifetime, 1 in 8 women will be diagnosed with breast cancer. There are no well established ways to avoid breast cancer (as opposed to lung cancer for example) and in the context of breast cancer screening, abnormalities should be detected at an early stage to improve prognosis. This project is directly relevant to human cancer since one of the end goals is to develop a computational method that would better assess tumor morphology and characterize disruptions in breast tissue microenvironment. The study of occult tumors and their microenvironment has become a hot cancer research topic recently. While most studies are done at the microscopic level and at fixed time points, the innovative aspects of this proposal are to establish a computational assessment methodology that would detect and characterize the disruption at the mammographic scale, and longitudinally. The technological implementation of this concept could significantly improve early cancer diagnostics.
The focus of this grant is to test the hypotheses that tissue disruption in the microenvironment of breast tumors may precede tumor apparition and development, and that the evolving morphology of the tumor itself may be intimately linked to its potential malignancy. We will computationally analyze a longitudinal set of mammograms and pathology reports of three groups of patients: those who have eventually had a diagnosis of cancer, those with benign tumors, and a control group. This will allow us to investigate the dynamic evolution of the morphology of the tumor as well as the changing landscape of its microenvironment. Verification of these hypotheses will lead to a better understanding of tumor apparition and growth, and to the development of a computational breast tissue assessment methodology that could determine when and where a tumor would eventually appear.
Since their FDA approval in 1998, increasing efforts have been made to develop novel and more accurate computer-aided diagnostic (CAD) methods. Unfortunately, several studies demonstrate that CAD methods are not offering the expected performance. Use of CAD in screening mammography is associated with decreased specificity and higher unnecessary recall rates. Additionally, the plethora of image analysis methods that have been developed as potential CAD methods have further limitations in that they are exclusively concentrated on tumor detection / characterization and they provide no link between the method's quantitative output and the biophysics of the underlying system. In contrast, we wish to open a new door in the computational analysis of mammograms by investigating the fractal vs. Euclidean morphology of breast lesions, as well as the disruption of the architecture of the breast tissue composing the microenvironment in the neighborhood of tumors, which, as we hypothesize, may precede the apparition of detectable tumors. We will use a wavelet-based image analysis method that will allow us to investigate, as a function of time, the evolution of the biophysics of breast lesions and their environment, in its healthy (organized, coherent) vs. unhealthy (disrupted, randomized) states.