Alex Dils Computer Vision · Medical AI
Vol. 03 / No. 02
Berkeley · 2026
§ Research

Computer vision for medical imaging, with an emphasis on robustness.

Two threads: reducing shortcut learning & confounding, and creating physiologically grounded augmentation for segmentation. Research intern at the Stanford Center for Biomedical Informatics Research since 2022.

01Themes

Three threads of work.

Thread A

Bias & confounders in diagnosis

Methods to reduce spurious correlations in image-based diagnosis and segmentation, and evaluation setups that surface failure modes beyond average performance.

Thread B

Physiologically informed augmentation

Simulation-driven perturbations that preserve anatomy / physics while expanding training support in clinically realistic ways.

Thread C

Segmentation & uncertainty

Practical approaches to improve segmentation robustness under shift, including augmentation design and careful metric & visual validation.

02Publications

Publications & manuscripts.

  1. Enhancing Medical AI with Physiologically-Informed Data Augmentation
    Christoph Sadée, Alex Dils, Cally Lin, Audrey Chun, Francisco Carrillo-Perez, Preya Shah, et al.
    NeurIPSsubmission in progress
    2026
  2. Addressing Bias and Confounders in AI-based Image Diagnosis — A Study of 117,610 Skin Lesions
    Christoph Sadée, Serena Zhang, Alex Dils, et al.
    Cellin revision
    Jan 2, 2025
  3. arXivpreprint
    Oct 27, 2024
    Read →
  4. Integrating Mechanistic Knowledge into Deep Learning for Improved Cancer Detection
    Christoph Sadée, Katherine Hartmann, Alex Dils, et al.
    FEniCS 2024proceedings
    Jun 16, 2024
  5. Eye For An Eye — A Deep-Learning & Analytical Method to Spatializing Stereoscopic Images
    Alex Dils*, Jake Yoshinaka, Leo McDonnell (*equal contribution)
    NHSJSpublication
    Apr 17, 2025

Full list and citations on Google Scholar.