Conformal Risk Control for Semantic Uncertainty
Quantification in Computed Tomography

This is the official implementation of the paper Conformal Risk Control for Semantic Uncertainty Quantification in Computed Tomography

by Jacopo Teneggi, J Webster Stayman, and Jeremias Sulam

---

Overview

sem-CRC measures the reconstruction uncertainty in semantic structures while ensuring coverage of the ground-truth image.

Furthermore, sem-CRC allows to control risk at the same level for each structure.

---

Usage

Install Dependencies

This project relies on the K-RCPS package available at Sulam-Group/k-rcps. To install it, run:

$ cd ..
$ git clone https://github.com/Sulam-Group/k-rcps.git
$ cd k-rcps
$ python -m pip install .

Download datasets: