All Times ET
Virtual
Statistical Pitfalls and Biases in Studies Using Biomedical Imaging (305267)
*Russell Shinohara, University of PennsylvaniaKeywords: Imaging science, biostatistics, biomedical imaging
Biomedical imaging is increasingly common in medical research, and can give unique insights into health and disease in vivo. However, conducting epidemiology and statistical analyses in imaging science settings pose new challenges and opportunities. Limitations of classical statistical tools, and important often under appreciated biases can threaten generalizability of findings. In this presentation, we will discuss common study designs, analytic approaches, and warning signs that are critical for the practice of statistics when using imaging data. We will also discuss new statistical tools, including software environments that allow for integrated image processing, to help address these issues and mitigate potential biases.