Time-Conditioned and Multi-Time Survival Prediction from 2D PET/CT Projections in Lung Cancer

2026-06-10 · Source: Takara TLDR - Daily AI Papers · Field: Science & Research — Health & Medical Research, Artificial Intelligence & Machine Learning, Mathematics & Computational Sciences · Depth: Expert, medium

Summary

A new study introduces Attention-guided Time-Conditioned Survival (ATCS) and Multi-Time Survival (MTS) models for predicting overall survival (OS) in non-small cell lung cancer (NSCLC) patients using pre-treatment 2D PET/CT projections. Researchers retrospectively analyzed 848 patient images, with 556 for development and 292 for testing, comparing the new models against a baseline Time-Conditioned Survival (TCS) model. Both ATCS and MTS significantly outperformed the baseline, achieving mean time-dependent AUCs of 0.794 and 0.793, respectively, compared to TCS's 0.767. ATCS showed superior performance for earlier predictions (0.5-3 years), while MTS excelled at later intervals (3.5-5 years). The study also found that combining tumor-specific and tissue-wise PET/CT features improved prediction accuracy, and temporal discretization influenced prediction horizons.

Key takeaway

For oncologists and AI scientists developing personalized treatment plans for NSCLC, integrating advanced temporal modeling into PET/CT analysis can significantly enhance overall survival predictions. You should consider implementing models like ATCS for more accurate short-term prognoses (0.5-3 years) and MTS for improved long-term estimates (3.5-5 years) to refine risk stratification and clinical decision-making.

Key insights

Temporal modeling significantly enhances PET/CT-based overall survival prediction in NSCLC patients.

Principles

• Temporal modeling improves survival prediction.
• Feature combination enhances model performance.
• Temporal discretization impacts prediction horizons.

Method

Developed Attention-guided Time-Conditioned Survival (ATCS) and Multi-Time Survival (MTS) models, trained with 5-fold cross-validation on PET/CT images, and evaluated using time-dependent AUC.

In practice

• Use ATCS for earlier survival predictions.
• Employ MTS for longer-term survival estimates.
• Combine tumor-specific and tissue-wise features.

Topics

• Lung Cancer
• Survival Prediction
• PET/CT Imaging
• Temporal Modeling
• Non-Small Cell Lung Cancer
• Machine Learning in Oncology

Best for: Computer Vision Engineer, AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Sybil: An AI model that can predict lung cancer risk 6 years in advance — Sybil AI predicts future lung cancer risk from CT scans by analyzing subtle, comprehensive image data.
• Traj-Evolve: A Self-Evolving Multi-Agent System for Patient Trajectory Modeling in Lung Cancer Early Detection — Traj-Evolve combines non-parametric memory and parametric reinforcement learning for robust patient trajectory modeling in complex EHR data.
• Virtual Scanning for NSCLC Histology: Investigating the Discriminatory Power of Synthetic PET — Synthetic PET data can significantly enhance NSCLC histological classification when combined with CT scans.
• I Gave a Model Eyes and a Genome. Here’s What It Learned About Brain Cancer. — Cross-attention fusion of pathology and genomics significantly improves glioblastoma survival prediction over unimodal approaches.
• Dynamic breast MRI with Flexible Temporal Resolution Aided by Deep Learning — Combining locally low-rank subspace modeling with deep learning enables flexible temporal resolution in DCE-MRI.
• Forecasting Medium-Horizon Alzheimer's Disease Progression: Residual Gap-Aware Transformers for 24-Month CDR-SB Change from ADNI Clinical and Biomarker Histories — Combining statistical anchoring with gap-aware residual learning improves medium-horizon Alzheimer's disease progression prediction.

Open in AIssential →

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