# Controlling for Biasing Signals in Images for Prognostic Models: Survival Predictions for Lung Cancer with Deep Learning

## TL;DR

This paper introduces a deep learning approach that uses causal modeling techniques to improve the accuracy and unbiasedness of lung cancer prognosis predictions from CT scans by accounting for confounding factors.

## Contribution

It proposes a novel method combining deep learning with causal inference to control for biasing signals in medical images for prognosis prediction.

## Key findings

- Unbiased prognosis predictions are achievable when the collider is quantified.
- Dual-task training improves the model's ability to account for confounding factors.
- Enforcing independence of last layer activations reduces bias in predictions.

## Abstract

Deep learning has shown remarkable results for image analysis and is expected to aid individual treatment decisions in health care. To achieve this, deep learning methods need to be promoted from the level of mere associations to being able to answer causal questions. We present a scenario with real-world medical images (CT-scans of lung cancers) and simulated outcome data. Through the sampling scheme, the images contain two distinct factors of variation that represent a collider and a prognostic factor. We show that when this collider can be quantified, unbiased individual prognosis predictions are attainable with deep learning. This is achieved by (1) setting a dual task for the network to predict both the outcome and the collider and (2) enforcing independence of the activation distributions of the last layer with ordinary least squares. Our method provides an example of combining deep learning and structural causal models for unbiased individual prognosis predictions.