Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AI-Generated Content Enhanced Computer-Aided Diagnosis Model for Thyroid Nodules: A ChatGPT-Style Assistant

Jincao Yao, Yunpeng Wang, Zhikai Lei, Kai Wang, Xiaoxian Li, Jianhua Zhou, Xiang Hao, Jiafei Shen, Zhenping Wang, Rongrong Ru, Yaqing Chen, Yahan Zhou, Chen Chen, Yanming Zhang, Ping Liang, Dong Xu

TL;DR

ThyGPT addresses the subjectivity and opacity of ultrasound-based thyroid nodule assessment by integrating a ChatGPT-style large language model with imaging analysis into an AIGC-CAD framework. Trained on a large, multi-source corpus including ultrasound images, anonymized reports, and international guidelines, it produces not only a malignancy probability but also a transparent rationale and feature attributions via heatmaps. In two independent test sets, ThyGPT enhanced radiologist performance, especially for junior clinicians, and demonstrated favorable standalone AI metrics, indicating potential to transform radiology workflows with explainable AI. Limitations include single-center data and inter-machine variability, highlighting the need for broader validation and iterative refinement.

Abstract

An artificial intelligence-generated content-enhanced computer-aided diagnosis (AIGC-CAD) model, designated as ThyGPT, has been developed. This model, inspired by the architecture of ChatGPT, could assist radiologists in assessing the risk of thyroid nodules through semantic-level human-machine interaction. A dataset comprising 19,165 thyroid nodule ultrasound cases from Zhejiang Cancer Hospital was assembled to facilitate the training and validation of the model. After training, ThyGPT could automatically evaluate thyroid nodule and engage in effective communication with physicians through human-computer interaction. The performance of ThyGPT was rigorously quantified using established metrics such as the receiver operating characteristic (ROC) curve, area under the curve (AUC), sensitivity, and specificity. The empirical findings revealed that radiologists, when supplemented with ThyGPT, markedly surpassed the diagnostic acumen of their peers utilizing traditional methods as well as the performance of the model in isolation. These findings suggest that AIGC-CAD systems, exemplified by ThyGPT, hold the promise to fundamentally transform the diagnostic workflows of radiologists in forthcoming years.

AI-Generated Content Enhanced Computer-Aided Diagnosis Model for Thyroid Nodules: A ChatGPT-Style Assistant

TL;DR

ThyGPT addresses the subjectivity and opacity of ultrasound-based thyroid nodule assessment by integrating a ChatGPT-style large language model with imaging analysis into an AIGC-CAD framework. Trained on a large, multi-source corpus including ultrasound images, anonymized reports, and international guidelines, it produces not only a malignancy probability but also a transparent rationale and feature attributions via heatmaps. In two independent test sets, ThyGPT enhanced radiologist performance, especially for junior clinicians, and demonstrated favorable standalone AI metrics, indicating potential to transform radiology workflows with explainable AI. Limitations include single-center data and inter-machine variability, highlighting the need for broader validation and iterative refinement.

Abstract

An artificial intelligence-generated content-enhanced computer-aided diagnosis (AIGC-CAD) model, designated as ThyGPT, has been developed. This model, inspired by the architecture of ChatGPT, could assist radiologists in assessing the risk of thyroid nodules through semantic-level human-machine interaction. A dataset comprising 19,165 thyroid nodule ultrasound cases from Zhejiang Cancer Hospital was assembled to facilitate the training and validation of the model. After training, ThyGPT could automatically evaluate thyroid nodule and engage in effective communication with physicians through human-computer interaction. The performance of ThyGPT was rigorously quantified using established metrics such as the receiver operating characteristic (ROC) curve, area under the curve (AUC), sensitivity, and specificity. The empirical findings revealed that radiologists, when supplemented with ThyGPT, markedly surpassed the diagnostic acumen of their peers utilizing traditional methods as well as the performance of the model in isolation. These findings suggest that AIGC-CAD systems, exemplified by ThyGPT, hold the promise to fundamentally transform the diagnostic workflows of radiologists in forthcoming years.
Paper Structure (10 sections, 5 figures)

This paper contains 10 sections, 5 figures.

Table of Contents

  1. Introduction
  2. Materials and Methods
  3. Patients
  4. Overall Design
  5. Model
  6. Image Acquisition
  7. Annotation and Preprocessing
  8. Statistical Analysis
  9. Results
  10. Discussion

Figures (5)

  • Figure 1: The overall design of our proposed ThyGPT model for thyroid nodules.
  • Figure 2: The internal design architecture of the proposed ThyGPT model.
  • Figure 3: Sample cases with physician errors and correct ThyGPT assessments.
  • Figure 4: Sample cases with physician correctness and erroneous ThyGPT assessments.
  • Figure 5: Clinical physicians' diagnostic results and ROC curves of deep learning methods.