What does this research mean for the field?

A few-shot prompting approach using Generative Pre-trained Transformer models, which refines predictions by incorporating examples of incorrect zero-shot predictions, outperforms BERT-based models and other open-source large language models in predicting lung cancer TNM staging from free-text radiology reports. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

To improve lung cancer clinical staging accuracy using natural language processing in radiology reports.

April 1, 2026Open Access

OURad at the NTCIR-18 RadNLP Task: Predicting Lung Cancer Clinical Staging from Radiology Reports Using Few-Shot Prompting of Large Language Models

Puntos clave

To improve lung cancer clinical staging accuracy using natural language processing in radiology reports.
Employed generative pre-trained transformer models for classification.
Utilized few-shot prompting approach for improved predictions.
Performed zero-shot prompting followed by refinement with incorrect examples.
Evaluated performance using a Japanese shared task on radiology reports.
Achieved a joint accuracy (fine) of 0.732 for the main task.
Obtained an overall micro F2.0 of 0.688 for the sub task.
Ranked 3rd in both the main task and sub task categories.

Resumen

In this paper, we describe our proposed systems for the Japanese main task and sub task in Natural Language Processing for Radiology 2024 shared task. We employed Generative Pre-trained Transformer models and applied a few-shot prompting approach to tackle the classification task for lung cancer TNM staging from free-text radiology reports. Our method first performs zero-shot prompting using training data and then refines the final predictions by incorporating examples of incorrect predictions into the prompt. We demonstrate that this approach outperforms several BERT-based models and other open-source large language models. On the test data, our method achieved a Joint Accuracy (fine) of 0.732 for the main task and an overall micro F2.0 of 0.688 for the sub task, ranking 3rd in both categories.

Leer artículo completoexternamente

Preguntar a la IA

Me gusta

Guardar

Ver artículo completo