analiza aiqbench
This study demonstrates that transformer models can effectively leverage longitudinal laboratory trajectories to predict treatment-associated complications in cancer patients weeks to months before clinical onset. By utilizing existing routine data, the approach achieves significant performance gains over single-timepoint baselines while generalizing across different healthcare systems without requiring new testing infrastructure. The findings highlight the potential for AI to extract hidden physiological signals from standard clinical records to enable proactive, complication-specific surveillance.
Routine laboratory trajectories encode the onset of organ-level complications in cancer
arXiv:2606.08538v1 Announce Type: new Abstract: Routine laboratory panels drawn during cancer treatment constitute longitudinal physiological recordings of organ function, yet their temporal structure is discarded by single-timepoint prognostic tools. A transformer trained on 2,777,595 laboratory measurements from 3,905 patients with multiple myeloma or ovarian cancer predicted the two-year onset of 162 treatment-associated complications, including therapy-related myelodysplastic syndromes, spanning eight clinical categories, achieving 1.5- to 6.1-fold enrichment above prevalence at the group level. It matched or outperformed non-sequential baselines across grouped endpoints (AUROC gains up to +0.11), demonstrating that longitudinal laboratory trajectories capture evolving complication-specific physiology inaccessible from isolated measurements. Predictions generalised across both cancers, divergence concentrating in disease-specific complications, and biomarker masking recovered signatures consistent with established pathophysiology. External validation on MIMIC-IV and MMRF CoMMpass confirmed transferability across independent healthcare systems (AUROC up to 0.85). Routine oncological laboratory data encode organ deterioration weeks to months before clinical onset, enabling complication-specific surveillance without additional testing infrastructure.
Abstract:Routine laboratory panels drawn during cancer treatment constitute longitudinal physiological recordings of organ function, yet their temporal structure is discarded by single-timepoint prognostic tools. A transformer trained on 2,777,595 laboratory measurements from 3,905 patients with multiple myeloma or ovarian cancer predicted the two-year onset of 162 treatment-associated complications, including therapy-related myelodysplastic syndromes, spanning eight clinical categories, achieving 1.5- to 6.1-fold enrichment above prevalence at the group level. It matched or outperformed non-sequential baselines across grouped endpoints (AUROC gains up to +0.11), demonstrating that longitudinal laboratory trajectories capture evolving complication-specific physiology inaccessible from isolated measurements. Predictions generalised across both cancers, divergence concentrating in disease-specific complications, and biomarker masking recovered signatures consistent with established pathophysiology. External validation on MIMIC-IV and MMRF CoMMpass confirmed transferability across independent healthcare systems (AUROC up to 0.85). Routine oncological laboratory data encode organ deterioration weeks to months before clinical onset, enabling complication-specific surveillance without additional testing infrastructure.
Machine Learning (cs.LG)
arXiv:2606.08538 [cs.LG]
(or arXiv:2606.08538v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2606.08538 arXiv-issued DOI via DataCite (pending registration)
Text integral sub licența deschisă a publicației · 168 cuvinte
Subiecte conexe
Valid Early Detection of Memorization in Diffusion Models
arXiv · 27 oct. 2026
Observation-Anchored Selective Assimilation for Longitudinal Tumor-State Proxy Forecasting in Post-Treatment Glioma
arXiv · 14 sept. 2026
Long-Horizon Language Model Reinforcement Learning via Progressive Point Matching
arXiv · 9 sept. 2026
Visual Prompting for Robotic Manipulation with Annotation-Guided Pick-and-Place Using ACT
arXiv · 19 aug. 2026