PRET is a few-shot system for pan-cancer recognition without example training

2026-04-03 · Source: Machine learning : nature.com subject feeds · Field: Health & Wellbeing — Clinical Care & Medical Practice, Medical Devices & Health Technology, Health & Medical Research · Depth: Expert, extended

Summary

A new few-shot AI system named PRET (pan-cancer recognition without examples training) has been developed to automate cancer diagnostics across various organs, hospitals, and tasks without requiring extensive training data. Published in Nature Cancer on April 3, 2026, PRET was evaluated on 23 international benchmarks, encompassing 4,484 whole-slide images. The system surpassed existing methods in 20 tasks, achieving over 97% area under the curve on 15 benchmarks, with a maximum improvement of 36.76%. Notably, PRET demonstrated clinical-grade diagnostic performance for lymph node metastasis detection using only eight slide examples, outperforming 11 human pathologists. This system offers a flexible and cost-effective solution, aiming to enhance accessibility and equity in AI-based pathology, particularly for underserved populations.

Key takeaway

For Computer Vision Engineers developing diagnostic AI, PRET demonstrates that few-shot learning can achieve clinical-grade performance in pan-cancer recognition with significantly less data than traditional methods. You should explore integrating visual in-context learning and patch-level feature processing into your models to enhance scalability and reduce training data requirements, especially for applications in underserved regions or with rare disease datasets.

Key insights

PRET enables flexible, scalable, and effective pan-cancer recognition with minimal examples, outperforming traditional AI and human pathologists.

Principles

• Few-shot learning reduces data dependency.
• Local patch-level features are crucial for accuracy.
• Visual in-context learning enhances model performance.

Method

PRET utilizes an extractor, tagger, miner, classifier, aggregator, and post-processor to leverage patch-level features and visual in-context prompts for training-free pan-cancer recognition.

In practice

• Deploy AI for cancer screening with limited data.
• Improve lymph node metastasis detection accuracy.
• Utilize visual prompts for flexible diagnostic tasks.

Topics

• PRET System
• Pan-Cancer Recognition
• Few-Shot Learning
• Digital Pathology
• Whole-Slide Images

Code references

• superjamessyx/robustness_benchmark
• xmed-lab/PRET

Best for: Computer Vision Engineer, AI Scientist, Research Scientist, Domain Expert

Related on AIssential

• Beyond Morphology: Quantifying the Diagnostic Power of Color Features in Cancer Classification — Pixel intensity alone, without morphology, provides a strong diagnostic signal for cancer classification.
• 😺 Mayo's AI spotted cancer 3 years before doctors did — AI is demonstrating superior diagnostic and triage capabilities in medicine, even with incomplete information.
• RARE disease detection from Capsule Endoscopic Videos based on Vision Transformers — Vision Transformers can be fine-tuned for multi-label classification of gastrointestinal diseases from capsule endoscopic videos.
• BreastGPT: A Multimodal Large Language Model for the Full Spectrum of Breast Cancer Clinical Routine — Workflow-aligned data and cross-scale visual modeling are critical for clinically grounded medical MLLMs.
• JASMINE — Early skin cancer detection in rural areas can be significantly improved with low-cost, offline, edge AI devices.
• A 3D SAM-Based Progressive Prompting Framework for Multi-Task Segmentation of Radiotherapy-induced Normal Tissue Injuries in Limited-Data Settings — A 3D SAM-based framework improves radiotherapy injury segmentation using progressive prompting and a small-target focus loss.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.