Auditing Sybil: Explaining Deep Lung Cancer Risk Prediction Through Generative Interventional Attributions

TL;DR

This paper introduces S(H)NAP, a causal auditing framework for deep lung cancer risk models like Sybil, revealing both its expert-like behavior and critical failure modes through generative interventional attributions.

Contribution

It presents the first interventional audit method for Sybil, using generative modeling to validate and analyze the model's causal reasoning in lung cancer risk prediction.

Findings

Sybil often mimics expert radiologist behavior in identifying malignant nodules.

The model shows dangerous sensitivity to clinically unjustified artifacts.

A radial bias was identified in the model's predictions.

Abstract

Lung cancer remains the leading cause of cancer mortality, driving the development of automated screening tools to alleviate radiologist workload. Standing at the frontier of this effort is Sybil, a deep learning model capable of predicting future risk solely from computed tomography (CT) with high precision. However, despite extensive clinical validation, current assessments rely purely on observational metrics. This correlation-based approach overlooks the model's actual reasoning mechanism, necessitating a shift to causal verification to ensure robust decision-making before clinical deployment. We propose S(H)NAP, a model-agnostic auditing framework that constructs generative interventional attributions validated by expert radiologists. By leveraging realistic 3D diffusion bridge modeling to systematically modify anatomical features, our approach isolates object-specific causal contributions to the risk score. Providing the first interventional audit of Sybil, we demonstrate that while the model often exhibits behavior akin to an expert radiologist, differentiating malignant pulmonary nodules from benign ones, it suffers from critical failure modes, including dangerous sensitivity to clinically unjustified artifacts and a distinct radial bias.