---
title: "Learning from Lost Provenance: Multiple Instance Learning for Cancer Registry Tumor Group Classification"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.03481v1"
permalink: "https://feed.astralor.com/items/itm_6ed08d3fdfbcb474cc13d122"
published: "2026-07-03T16:48:06.000Z"
score: 77
channel: "tech"
tags: ["Attention-Based Multiple Instance Learning","BC Cancer Registry","Weakly Supervised Learning","Deep Learning","Pathology Report Coding"]
---

# Learning from Lost Provenance: Multiple Instance Learning for Cancer Registry Tumor Group Classification

## AI commentary

> This paper catches my eye by cleverly sidestepping the 'expensive annotation' trap in medical AI, instead leveraging existing coarse labels from hospital operations. It's a treasure for any institution looking to reduce operational costs with AI but lacking annotation resources. Particularly valuable for pathology departments, cancer registries, and researchers in weakly supervised learning.

## At a glance

Modernizing cancer registries with deep learning for pathology report coding is hampered by the lack of report-level annotations. Researchers at BC Cancer Registry leverage patient-level operational labels via Attention-Based Multiple Instance Learning (ABMIL) to recover the lost link between labels and reports. Distilling a noisy corpus yields a compact training set, achieving macro F1 of 0.83, outperforming baselines. This framework requires no extra annotation or large-scale computing, offering a practical route to automate cancer registry workflows.

## Key points

- ABMIL uses existing patient-level operational labels from cancer registries, eliminating need for report-level annotation.
- Attention mechanism selects the most informative pathology reports to distill a high-quality training set.
- Achieves macro F1 of 0.83 on BC Cancer Registry tumor group classification, surpassing multiple baselines.
- Low computational cost and easy to deploy, reducing burden on human coders.

## Full text

Modernizing cancer registries with deep learning is opening new opportunities to automate labor-intensive tasks such as the coding of pathology reports. However, progress is constrained by the scarcity of report-level human-annotated training data. Cancer registries generate substantial volumes of expert-assigned labels as a routine product of their operations, but these exist at the patient level and are not linked to the individual pathology reports that informed them, limiting their direct use for training models. We develop an efficient framework for training deep learning classifiers by leveraging these operationally-generated labels without requiring per-report human annotation, demonstrated for tumor group classification at the BC Cancer Registry. We use Attention-Based Multiple Instance Learning (ABMIL) to recover the lost link between patient-level labels and the reports that informed them, leveraging the attention the model places on each report to distil a large, noisily-labeled corpus into a compact, high-quality per-report training dataset. A classifier fine-tuned on a distilled dataset achieved a macro F1 of 0.83, outperforming established baselines across most tumor groups. By turning routine operational labels into high-quality training data without additional annotation or large-scale computing infrastructure, ABMIL offers a practical and accessible route to automating cancer registry workflows.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_6ed08d3fdfbcb474cc13d122