MOOZY: A Patient-First Foundation Model for Computational Pathology

Abstract

Computational pathology needs whole-slide image (WSI) foundation models that transfer across diverse clinical tasks, yet current approaches remain largely slide-centric, often depend on private data and expensive paired-report supervision, and do not explicitly model relationships among multiple slides from the same patient. We present MOOZY, a patient-first pathology foundation model in which the patient case, not the individual slide, is the core unit of representation. MOOZY explicitly models dependencies across all slides from the same patient via a case transformer during pretraining, combining multi-stage open self-supervision with scaled low-cost task supervision. In Stage 1, we pretrain a vision-only slide encoder on 77,134 public slide feature grids using masked self-distillation. In Stage 2, we align these representations with clinical semantics using a case transformer and multi-task supervision over 333 tasks from 56 public datasets, including 205 classification and 128 survival tasks across four endpoints. Across eight held-out tasks with five-fold frozen-feature probe evaluation, MOOZY achieves best or tied-best performance on most metrics and improves macro averages over TITAN by +7.37%, +5.50%, and +7.83% and over PRISM by +8.83%, +10.70%, and +9.78% for weighted F1, weighted ROC-AUC, and balanced accuracy, respectively. MOOZY is also parameter efficient with 85.77M parameters, 14x smaller than GigaPath. These results demonstrate that open, reproducible patient-level pretraining yields transferable embeddings, providing a practical path toward scalable patient-first histopathology foundation models.

Figures (17)

• Figure 1: (a) Weighted F1 across eight held-out tasks. Brackets report [min -- max] weighted F1 for each task, with the center corresponding to the minimum and the outer ring to the maximum observed value. (b) Macro-averaged weighted F1 versus total parameter count (log scale), showing that MOOZY remains highly accurate while being highly parameter-efficient.
• Figure 2: Overview of the proposed two-stage framework. Stage 1 (up): A frozen patch encoder extracts per-patch features arranged into a spatial grid. Multi-scale crops are sampled with spatial augmentations and block-based masking. A student slide encoder and EMA teacher are jointly trained via CLS-level self-distillation ($\mathcal{L}_\text{cls}$) and masked patch prediction ($\mathcal{L}_\text{mim}$). Stage 2 (down): The pretrained slide encoder produces per-slide embeddings; a case transformer aggregates them into a unified case embedding $\tilde{\mathbf{h}}_i$, routed to task-specific classification and survival heads.
• Figure 3: Architecture of the slide encoder and case aggregator. (A) The slide encoder takes patch embeddings, a learnable [CLS] token, $R$ register tokens, and mask tokens, processed through $D$ transformer blocks. (B) The case aggregator prepends a learnable [CASE] token to per-slide embeddings and produces a case embedding $\tilde{\mathbf{h}}_i$, routed to heads for classification and survival prediction.
• Figure 4: Radial hierarchy of MOOZY data scale across four dimensions: pretraining scale, anatomical coverage, task taxonomy, and supervision structure.
• Figure 5: Attention-map comparison on a lung adenocarcinoma slide. MOOZY and TITAN: balanced, comprehensive coverage (shift 3, gap 1). PRISM: balanced shift with moderate gaps (shift 3, gap 2). CHIEF and Madeleine: cancer-biased with frequent semantic gaps (shift 2, gap 3).