Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards Clinical Practice in CT-Based Pulmonary Disease Screening: An Efficient and Reliable Framework

Qian Shao, Bang Du, Yixuan Wu, Zepeng Li, Qiyuan Chen, Qianqian Tang, Jian Wu, Jintai Chen, Hongxia Xu

Abstract

Deep learning models for pulmonary disease screening from Computed Tomography (CT) scans promise to alleviate the immense workload on radiologists. Still, their high computational cost, stemming from processing entire 3D volumes, remains a major barrier to widespread clinical adoption. Current sub-sampling techniques often compromise diagnostic integrity by introducing artifacts or discarding critical information. To overcome these limitations, we propose an Efficient and Reliable Framework (ERF) that fundamentally improves the practicality of automated CT analysis. Our framework introduces two core innovations: (1) A Cluster-based Sub-Sampling (CSS) method that efficiently selects a compact yet comprehensive subset of CT slices by optimizing for both representativeness and diversity. By integrating an efficient k-nearest neighbor search with an iterative refinement process, CSS bypasses the computational bottlenecks of previous methods while preserving vital diagnostic features. (2) An Ambiguity-aware Uncertainty Quantification (AUQ) mechanism, which enhances reliability by specifically targeting data ambiguity arising from subtle lesions and artifacts. Unlike standard uncertainty measures, AUQ leverages the predictive discrepancy between auxiliary classifiers to construct a specialized ambiguity score. By maximizing this discrepancy during training, the system effectively flags ambiguous samples where the model lacks confidence due to visual noise or intricate pathologies. Validated on two public datasets with 2,654 CT volumes across diagnostic tasks for 3 pulmonary diseases, ERF achieves diagnostic performance comparable to the full-volume analysis (over 90% accuracy and recall) while reducing processing time by more than 60%. This work represents a significant step towards deploying fast, accurate, and trustworthy AI-powered screening tools in time-sensitive clinical settings.

Towards Clinical Practice in CT-Based Pulmonary Disease Screening: An Efficient and Reliable Framework

Abstract

Deep learning models for pulmonary disease screening from Computed Tomography (CT) scans promise to alleviate the immense workload on radiologists. Still, their high computational cost, stemming from processing entire 3D volumes, remains a major barrier to widespread clinical adoption. Current sub-sampling techniques often compromise diagnostic integrity by introducing artifacts or discarding critical information. To overcome these limitations, we propose an Efficient and Reliable Framework (ERF) that fundamentally improves the practicality of automated CT analysis. Our framework introduces two core innovations: (1) A Cluster-based Sub-Sampling (CSS) method that efficiently selects a compact yet comprehensive subset of CT slices by optimizing for both representativeness and diversity. By integrating an efficient k-nearest neighbor search with an iterative refinement process, CSS bypasses the computational bottlenecks of previous methods while preserving vital diagnostic features. (2) An Ambiguity-aware Uncertainty Quantification (AUQ) mechanism, which enhances reliability by specifically targeting data ambiguity arising from subtle lesions and artifacts. Unlike standard uncertainty measures, AUQ leverages the predictive discrepancy between auxiliary classifiers to construct a specialized ambiguity score. By maximizing this discrepancy during training, the system effectively flags ambiguous samples where the model lacks confidence due to visual noise or intricate pathologies. Validated on two public datasets with 2,654 CT volumes across diagnostic tasks for 3 pulmonary diseases, ERF achieves diagnostic performance comparable to the full-volume analysis (over 90% accuracy and recall) while reducing processing time by more than 60%. This work represents a significant step towards deploying fast, accurate, and trustworthy AI-powered screening tools in time-sensitive clinical settings.

Paper Structure

This paper contains 32 sections, 6 equations, 4 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Pulmonary Disease Screening
  4. Sub-Sampling
  5. Uncertainty Quantification
  6. Methods
  7. Cluster-based Sub-Sampling
  8. Partitioning and Representation
  9. Iterative Selection via Density and Diversity
  10. Representativeness
  11. Diversity Refinement
  12. Selection Objective
  13. Ambiguity-aware Uncertainty Quantification (AUQ)
  14. Framework
  15. Discrepancy Maximization
  16. ...and 17 more sections

Figures (4)

  • Figure 1: Visualization of interpolation methods on common pneumonia (CP) lesions. (a-d) are four consecutive slices showing CP lesions (red solid-line area). (e) and (f) are results of spline interpolation and projection interpolation, respectively. Yellow solid lines indicate processing-induced artifacts, blue dashed lines highlight regions of lesion elimination post-processing.
  • Figure 2: The core insight of CSS. Density peaks are selected for representativeness, and then they are pushed away via an iterative refinement process for diversity. The dashed and solid black circles denote the selected instances from the previous iteration and the current iteration, respectively.
  • Figure 3: Illustration of AUQ. The circles of different colors represent samples from distinct categories. The solid and dashed lines denote the decision hyperplanes generated by the main and auxiliary classifiers, respectively. High-uncertainty samples are located in the gray area.
  • Figure 4: CT slices selected by CSS. The red, yellow, and blue boxes mark the lesion areas of NCP, CP, and AC, respectively.