Imaging, Data, and Learning:

Jeremias Sulam

Modern challenges in biomedical datadata science

ISPEED 2025

Inverse Problems

Micro


Meso
Macro

Imaging, Data, and Learning

Fairness in Data Science

Is the model fair?

\text{TPR}_\text{sex} = \mathbb P [\hat Y = 1 | Y = 1, \text{sex}]

Pneumonia:  \(Y=1\)

Clear: \(Y=0\)

95% accurate

\Delta_{\text{TPR}} = {\Large|} \text{True Positives}_\text{males} - \text{True Positives}_\text{females} {\Large|}

Fairness in Data Science

predictions
ground-truth

Does your model achieve a fairness violation of at most (say) 6% ?

Fairness in Data Science

Pneumonia:  \(Y=1\)

Clear: \(Y=0\)

95% accurate

  • Tight upper bounds to fairness violations
     

  • (optimally) Actionable
     

Maximum TPR
discrepancy

True TPR
discrepancy

Bharti, B., Yi, P., & Sulam, J. (2023). Estimating and Controlling for Equalized Odds via Sensitive Attribute Predictors NeurIPS 2023

Fairness in Data Science

Pneumonia

Clear

95% accurate

Macro

Meso
Micro

Imaging, Data, and Learning

Micro

Imaging, Data, and Learning

Micro
Macro

Meso

Imaging, Data, and Learning

“The future rewards those who press on.”

B. Obama

Inverse Problems in Neuroscience: Susceptibility Tensor Imaging

6 times faster

Inverse Problems in Digital Pathology

Inverse Problems in Radiology

Opportunity is missed by most people because it is dressed in overalls and looks like work.

T. Edison