Uncertainty-Aware Prediction of Lung Tumor Growth from Sparse Longitudinal CT Data via Bayesian Physics-Informed Neural Networks

2026-05-13 · Source: Takara TLDR - Daily AI Papers · Field: Health & Wellbeing — Health & Medical Research, Medical Specialties & Subspecialties, Medical Devices & Health Technology · Depth: Expert, medium

Summary

A new Bayesian physics-informed neural network (BPINN) has been developed to predict lung tumor growth from sparse and irregular longitudinal computed tomography (CT) observations, accounting for measurement variability. The framework integrates Gompertz growth dynamics with low-dimensional Bayesian inference in the log-volume domain. It uses a two-stage inference strategy, combining maximum a posteriori (MAP) estimation and Hamiltonian Monte Carlo (HMC) sampling, to generate posterior predictive distributions and uncertainty intervals. Evaluated on longitudinal data from 30 patients in the National Lung Screening Trial, the model achieved a cohort-level log-space RMSE of approximately 0.20. It successfully captured heterogeneous tumor growth patterns and provided calibrated 95% credible interval coverage, outperforming deterministic methods by offering robust uncertainty estimates.

Key takeaway

For AI Scientists developing predictive models for medical applications, this work demonstrates that integrating Bayesian physics-informed neural networks with two-stage inference can yield accurate predictions and crucial uncertainty quantification from sparse longitudinal data. You should consider adopting this framework to enhance the reliability of your models, especially when dealing with limited patient follow-up scans, as it provides well-calibrated credible intervals essential for clinical decision support.

Key insights

BPINNs can predict lung tumor growth with calibrated uncertainty from sparse CT data.

Principles

• Combine Gompertz dynamics with Bayesian inference.
• Use two-stage inference for posterior estimation.
• Calibrated uncertainty improves tumor growth assessment.

Method

A two-stage inference strategy combines MAP estimation and HMC sampling to estimate posterior predictive distributions and uncertainty intervals for tumor growth parameters within a Bayesian physics-informed neural network.

In practice

• Apply BPINN for sparse longitudinal medical data.
• Use MAP and HMC for robust uncertainty.
• Assess tumor growth with calibrated uncertainty.

Topics

• Bayesian Physics-Informed Neural Networks
• Lung Tumor Growth Prediction
• Uncertainty Quantification
• Sparse Longitudinal CT Data
• Gompertz Growth Dynamics

Best for: AI Scientist, Research Scientist

Related on AIssential

• Neural network surrogates with uncertainty quantification for inverse problems in partial differential equations — DeepGaLA uses uncertainty-aware neural network surrogates for scalable, reliable Bayesian inference in inverse problems, outperforming traditional methods in efficiency.
• Uncertainty-Aware Longitudinal Forecasting of Alzheimer's Disease Progression Using Deep Learning — A probabilistic deep learning framework forecasts Alzheimer's disease progression with quantified uncertainty, outperforming existing baselines.
• When is 3D Worth It? A Resource-Performance Frontier for CNNs and Transformers in Lung CT — Higher dimensionality in medical imaging models does not guarantee stable performance, especially with class imbalance.
• I got frustrated teaching ML to scientists, so I started building domain-specific workshops – would love your thoughts — Standard ML education often fails to address the unique data constraints and uncertainty requirements of scientific research.
• Bayesian Magnetic Resonance Joint Image Reconstruction and Uncertainty Quantification using Sparsity Prior Models and Markov Chain Monte Carlo Sampling — Bayesian MCMC with sparsity priors enables superior MR image reconstruction and uncertainty quantification, outperforming optimization and deep learning methods.
• A Streaming Sparse Cholesky Method for Derivative-Informed Gaussian Process Surrogates Within Digital Twin Applications — Derivative-informed sparse Gaussian Processes enable accurate, dynamically updated digital twin surrogates for real-time system prognostics.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.