Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Personalized Longitudinal Medical Report Generation via Temporally-Aware Federated Adaptation

He Zhu, Ren Togo, Takahiro Ogawa, Kenji Hirata, Minghui Tang, Takaaki Yoshimura, Hiroyuki Sugimori, Noriko Nishioka, Yukie Shimizu, Kohsuke Kudo, Miki Haseyama

TL;DR

FTA is introduced, a federated setting that explicitly accounts for the temporal evolution of client data and FedTAR is proposed, a framework that integrates demographic-driven personalization with time-aware global aggregation and demonstrates consistent improvements in linguistic accuracy, temporal coherence, and cross-site generalization.

Abstract

Longitudinal medical report generation is clinically important yet remains challenging due to strict privacy constraints and the evolving nature of disease progression. Although federated learning (FL) enables collaborative training without data sharing, existing FL methods largely overlook longitudinal dynamics by assuming stationary client distributions, making them unable to model temporal shifts across visits or patient-specific heterogeneity-ultimately leading to unstable optimization and suboptimal report generation. We introduce Federated Temporal Adaptation (FTA), a federated setting that explicitly accounts for the temporal evolution of client data. Building upon this setting, we propose FedTAR, a framework that integrates demographic-driven personalization with time-aware global aggregation. FedTAR generates lightweight LoRA adapters from demographic embeddings and performs temporal residual aggregation, where updates from different visits are weighted by a meta-learned temporal policy optimized via first-order MAML. Experiments on J-MID (1M exams) and MIMIC-CXR demonstrate consistent improvements in linguistic accuracy, temporal coherence, and cross-site generalization, establishing FedTAR as a robust and privacy-preserving paradigm for federated longitudinal modeling.

Personalized Longitudinal Medical Report Generation via Temporally-Aware Federated Adaptation

TL;DR

FTA is introduced, a federated setting that explicitly accounts for the temporal evolution of client data and FedTAR is proposed, a framework that integrates demographic-driven personalization with time-aware global aggregation and demonstrates consistent improvements in linguistic accuracy, temporal coherence, and cross-site generalization.

Abstract

Longitudinal medical report generation is clinically important yet remains challenging due to strict privacy constraints and the evolving nature of disease progression. Although federated learning (FL) enables collaborative training without data sharing, existing FL methods largely overlook longitudinal dynamics by assuming stationary client distributions, making them unable to model temporal shifts across visits or patient-specific heterogeneity-ultimately leading to unstable optimization and suboptimal report generation. We introduce Federated Temporal Adaptation (FTA), a federated setting that explicitly accounts for the temporal evolution of client data. Building upon this setting, we propose FedTAR, a framework that integrates demographic-driven personalization with time-aware global aggregation. FedTAR generates lightweight LoRA adapters from demographic embeddings and performs temporal residual aggregation, where updates from different visits are weighted by a meta-learned temporal policy optimized via first-order MAML. Experiments on J-MID (1M exams) and MIMIC-CXR demonstrate consistent improvements in linguistic accuracy, temporal coherence, and cross-site generalization, establishing FedTAR as a robust and privacy-preserving paradigm for federated longitudinal modeling.
Paper Structure (45 sections, 7 theorems, 47 equations, 6 figures, 6 tables, 2 algorithms)

This paper contains 45 sections, 7 theorems, 47 equations, 6 figures, 6 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Federated Learning for Medical Applications
  4. Temporal Longitudinal Medical Imaging
  5. Methodology
  6. FTA Problem Definition
  7. Preliminary
  8. Client-Side Patient-Adaptive Updates
  9. Temporal Residual Aggregation
  10. Two Layer Temporal Aggregation
  11. Meta-Learned Coefficients
  12. Experiments
  13. Dataset Composition
  14. Federated Client Partitioning
  15. Ethics and Privacy
  16. ...and 30 more sections

Key Result

Theorem 1

Let $\Delta_t = \bar{\bm w}_g^{(t,r)} - \bm w_g^{(t-1,r)}$, $\beta_x^{(t)} = \alpha_x \prod_{y=x+1}^t (1 - \alpha_y)\quad (1\le x\le t)$, and $\beta_0^{(t)} = \prod_{y=1}^t (1 - \alpha_y).$ Then the iterates defined by eq:residual_update admit the closed form with $\beta_x^{(t)}\ge0$ and $\sum_{x=0}^t\beta_x^{(t)}=1$. Consequently, $\bm w_g^{(t,r)}$ always lies in the convex hull of $\{\bar{\bm w

Figures (6)

  • Figure 1: Visualization of semantic structure evolution in medical reports. All data are derived from the J-MID dataset. (a) Ridgeline kernel-density plots of LDA-derived topic components for annual reports (2018–2024) from one institution. (b) T-SNE embedding of TF-IDF vectors for reports from five institutions.
  • Figure 2: Overview of the problem and our approach. (a) Client- and temporal-drift challenges in federated longitudinal medical report generation. (b) The proposed FedTAR framework.
  • Figure 3: Overall caption describing both subfigures.
  • Figure 4: fig:generated report
  • Figure :
  • ...and 1 more figures

Theorems & Definitions (12)

  • Theorem 1
  • Theorem 2
  • Lemma 3
  • proof
  • Lemma 4
  • proof
  • Lemma 5: Exact increment expression
  • proof
  • Lemma 6: Norm bound for a scaled residual
  • proof
  • ...and 2 more