MICCAI 2021 Daily – Wednesday

Many clinically useful questions are counterfactuals. For example, would a patient’s lesion burden respond better to drug A or drug B? This paper proposes a model using ideas from causal inference to generate counterfactual MRI images that imagine what a scan of someone with Multiple Sclerosis would look like if they did not have MS, which is useful for disease modeling and other technical applications. 12 DAILY MICCAI Wednesday Poster Presentation A Structural Causal Model MR Images of Multiple Sclerosis Jacob Reinhold is a data scientist at the Memorial Sloan Kettering Cancer Center. In his previous role as a graduate research assistant at Johns Hopkins University, under the supervision of Jerry Prince, he proposed an approach to generating counterfactual images of MRIs for people with multiple sclerosis (MS). He speaks to us about this work ahead of his poster session today. In the picture, you can see two examples of counterfactual images of a patient with MS; one image shows the counterfactual with the ventricle volume set to 80mL and the other shows the counterfactual with the lesion volume set to 0mL.