Beyond the P-Value
Weekday Course
ORGANIZERS: Jana Delfino, Shella Keilholz
Tuesday, 13 May 2025
320
13:30 -
15:30
Moderators: Cristian Montalba & Tony Stoecker
Skill Level: Basic to Intermediate
Session Number: Tu-02
No CME/CE Credit
Session Number: Tu-02
Overview
Learning statistics is of significant importance among different professionals that work in radiology research, such as data interpretation, study design, hypothesis testing, ensuring data quality and outcome prediction. There are several statistical techniques that are applied depending on the data type, number of groups and observations. Additionally, choosing the right statistical test to demonstrate robustness of your findings can be challenging. This tutorial will be an interactive session to provide guidance on how to properly present the results of methodological studies in a paper, so that the presentation is clear, confidence intervals are properly identified, and the conclusions are statistically justified. Examples of 1) the characterization of a quantitative imaging method, and 2) a case-control study to compare an MRI metric between clinical groups will be presented through jupyter notebooks that the participants can follow and experiment upon during the session itself. The session will present basic statistical concepts and tools, such as sample size, confidence intervals, when to use or not to use inferential statistics and significance testing.
Target Audience
Students and researchers from healthcare and engineering fields that use statistical testing.
Educational Objectives
As a result of attending this course, participants should be able to:
• Summarize how and when to use common statistical tests;
• Explain what common statistical tests can and cannot show;
• Determine correct (and incorrect) use of statistical testing; and
• Apply statistical testing in practice.
| 13:30 | | Introduction: Why Statistics Matters to Us All Sophie Schauman Keywords: Transferable skills: Statistics Statistics are of significant importance to all professionals working on MR related research. This introductory talk will show how using the correct statistical methods and interpretations is crucial throughout the research cycle, whatever the focus of the research is. The presentation will highlight examples of what can happen when statistics are used incorrectly, as well as how correct application of statistical methods can improve rigor and confidence in research findings, leading to increased scientific impact. |
| 14:00 | | The Characterization of a Quantitative Imaging Method Francesco Santini Keywords: Transferable skills: Statistics, Transferable skills: Reproducible research, Image acquisition: Quantification You have developed a new quantitative imaging method. Congratulations! Now what? How can you verify if this method is as good as you wish it to be? In this talk, we will see how to use statistics to properly and convincingly characterize your results, and fairly evaluate your novel method. Following the guidance of the consensus paper by the Quantitative MRI Study Group published in 2021, we will delve deeper into what the most important metrics are, how to rigorously calculate them, and what kind of sample sizes are required, depending on the diagnostic requirements. |
| 14:30 | | A Case-Control Study To Compare an MRI Metric Between Clinical Groups Maria Eugenia Caligiuri Keywords: Transferable skills: Reproducible research You are planning to investigate differences in MRI-derived measures between patients and healthy subjects, or across different subtypes of a disease or syndrome. That’s great! But what is the appropriate analysis path you should follow? Is traditional statistics appropriate and sufficient, or is it worth trying artificial intelligence (it’s so mainstream nowadays!!)? In this talk, we’ll show you how to address this question considering the nature of your dataset and the scientific hypothesis you want to test (describe disease-related tissue alterations, diagnosis prediction, etc.). |