Dr. Jose-Miguel Yamal
The University of Texas Health Science Center at Houston
School of Public Health
My primary research interests are: (1) statistical methodology for classification of high-dimensional data (e.g., hierarchical and functional data), (2) clinical trials, with a focus on diagnostic test studies, and (3) automated detection of cancer (cervix and breast). We have contributed to the statistical methodology of classification when the data structure is hierarchical, e.g., classifying patients using cell-level high-dimensional data (Yamal, 2011). We have also conducted studies of functional data, mainly tissue fluorescence measurements for the detection of cervical cancer, where the unit of observation is a matrix of intensities at several excitation-emission pair wavelengths of light. We have explored ways to use this functional data for prediction of disease (Cantor, 2011 and Yamal, 2012) and have proposed a method to assess repeatability of functional data (Yamal, 2010).
I am the Principal Investigator of the Statistical Core for the National Institutes of Health (NIH) sponsored clinical trial Effects of Erythropoietin on Cerebral Vascular Dysfunction & Anemia in Traumatic Brain Injury. I am also working on two diagnostic test studies: (1) a study developing, training and testing low-cost technologies for the detection of cervical cancer and (2) a study using elastography and novel algorithms and features to discriminate between benign and malignant breast tumors.
The use of statistical methods and algorithms for disease detection and discrimination is my third main research interest. The current standard of care in the developed world for cervical cancer screening suffers from high inter/intra-provider variability, low sensitivity, the high expense of manpower. In the developing world, there is no current standard of care. I, with collaborators at UT M.D. Anderson Cancer Center, Drexel University, Texas Tech, British Columbia Cancer Agency, and in Nigeria, have been researching several technologies to address this problem: quantitative cytology (Yamal 2004, Malpica 2005, Kim 2005, Yamal 2011, and forthcoming manuscripts), fluorescence spectroscopy (Cantor 2011, Yamal 2012), and a combined multispectral digital colposcope (MDC) and fluorescence spectroscopy device (in an upcoming study of 650 patients). Using statistical models, we have been developing and testing algorithms to automate the detection of cervical cancer using data from these devices.
A tutorial in my laboratory could provide experience with classification methods, diagnostic trials, diagnostic device development, and cervical and breast cancer detection.