Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LMM-IQA: Image Quality Assessment for Low-Dose CT Imaging

Kagan Celik, Mehmet Ozan Unal, Metin Ertas, Isa Yildirim

TL;DR

This work tackles image quality assessment for low-dose CT by leveraging large multimodal models to produce both numerical scores and textual explanations of LDCT degradations. It systematically evaluates a progressive inference framework—from zero-shot to metadata-enhanced and error-feedback–guided inference—across a large LDCT-IQA dataset, using PLCC, SROCC, and KROCC as metrics and summing them for an Overall Score. Key findings show that zero-shot performance is weak, small-scale few-shot improves significantly, and incorporating region/noise metadata as well as error-feedback yields the strongest correlations with radiologist judgments (best Overall Score ≈ 2.26). The approach offers valuable interpretability and potential clinical utility by combining quantitative scores with descriptive quality factors, suggesting a flexible, training-free path for integrating LMMs into LDCT quality assessment and decision support.

Abstract

Low-dose computed tomography (CT) represents a significant improvement in patient safety through lower radiation doses, but increased noise, blur, and contrast loss can diminish diagnostic quality. Therefore, consistency and robustness in image quality assessment become essential for clinical applications. In this study, we propose an LLM-based quality assessment system that generates both numerical scores and textual descriptions of degradations such as noise, blur, and contrast loss. Furthermore, various inference strategies - from the zero-shot approach to metadata integration and error feedback - are systematically examined, demonstrating the progressive contribution of each method to overall performance. The resultant assessments yield not only highly correlated scores but also interpretable output, thereby adding value to clinical workflows. The source codes of our study are available at https://github.com/itu-biai/lmms_ldct_iqa.

LMM-IQA: Image Quality Assessment for Low-Dose CT Imaging

TL;DR

This work tackles image quality assessment for low-dose CT by leveraging large multimodal models to produce both numerical scores and textual explanations of LDCT degradations. It systematically evaluates a progressive inference framework—from zero-shot to metadata-enhanced and error-feedback–guided inference—across a large LDCT-IQA dataset, using PLCC, SROCC, and KROCC as metrics and summing them for an Overall Score. Key findings show that zero-shot performance is weak, small-scale few-shot improves significantly, and incorporating region/noise metadata as well as error-feedback yields the strongest correlations with radiologist judgments (best Overall Score ≈ 2.26). The approach offers valuable interpretability and potential clinical utility by combining quantitative scores with descriptive quality factors, suggesting a flexible, training-free path for integrating LMMs into LDCT quality assessment and decision support.

Abstract

Low-dose computed tomography (CT) represents a significant improvement in patient safety through lower radiation doses, but increased noise, blur, and contrast loss can diminish diagnostic quality. Therefore, consistency and robustness in image quality assessment become essential for clinical applications. In this study, we propose an LLM-based quality assessment system that generates both numerical scores and textual descriptions of degradations such as noise, blur, and contrast loss. Furthermore, various inference strategies - from the zero-shot approach to metadata integration and error feedback - are systematically examined, demonstrating the progressive contribution of each method to overall performance. The resultant assessments yield not only highly correlated scores but also interpretable output, thereby adding value to clinical workflows. The source codes of our study are available at https://github.com/itu-biai/lmms_ldct_iqa.

Paper Structure

This paper contains 17 sections, 5 equations, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Dataset
  4. Evaluated Large Language and Multimodal Models
  5. Proposed LMM-Based IQA Frameworks
  6. Zero-Shot Inference
  7. Few-Shot Inference (Full Dataset)
  8. Error Feedback–Guided Inference
  9. Metadata Integration for Context-Aware Scoring
  10. Evaluation Metrics
  11. Experimental Results
  12. Zero-Shot Inference Results
  13. Few-Shot Inference Results (Small Scale - 10 Images)
  14. Few-Shot Inference Results (Full Dataset – 1000 Training / 300 Test)
  15. Few-Shot Inference Results with Metadata
  16. ...and 2 more sections

Figures (3)

  • Figure 1: Flowchart of the LMM-IQA Methodology
  • Figure 2: Scatter Plot Comparing Radiologist Scores with O3-Error Feedback Scores
  • Figure 3: Distribution of Radiologist vs O3-Error Feedback Scores