Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Rad4XCNN: a new agnostic method for post-hoc global explanation of CNN-derived features by means of radiomics

Francesco Prinzi, Carmelo Militello, Calogero Zarcaro, Tommaso Vincenzo Bartolotta, Salvatore Gaglio, Salvatore Vitabile

TL;DR

This paper presents a novel method, namely Rad4XCNN, to enhance the predictive power of CNN-derived features with the inherent interpretability of radiomic features, by associating intelligible meaning to CNN-derived features by means of Radiomics, offering new perspectives on explanation methods beyond visualization maps.

Abstract

In recent years, machine learning-based clinical decision support systems (CDSS) have played a key role in the analysis of several medical conditions. Despite their promising capabilities, the lack of transparency in AI models poses significant challenges, particularly in medical contexts where reliability is a mandatory aspect. However, it appears that explainability is inversely proportional to accuracy. For this reason, achieving transparency without compromising predictive accuracy remains a key challenge. This paper presents a novel method, namely Rad4XCNN, to enhance the predictive power of CNN-derived features with the inherent interpretability of radiomic features. Rad4XCNN diverges from conventional methods based on saliency maps, by associating intelligible meaning to CNN-derived features by means of Radiomics, offering new perspectives on explanation methods beyond visualization maps. Using a breast cancer classification task as a case study, we evaluated Rad4XCNN on ultrasound imaging datasets, including an online dataset and two in-house datasets for internal and external validation. Some key results are: i) CNN-derived features guarantee more robust accuracy when compared against ViT-derived and radiomic features; ii) conventional visualization map methods for explanation present several pitfalls; iii) Rad4XCNN does not sacrifice model accuracy for their explainability; iv) Rad4XCNN provides a global explanation enabling the physician to extract global insights and findings. Our method can mitigate some concerns related to the explainability-accuracy trade-off. This study highlighted the importance of proposing new methods for model explanation without affecting their accuracy.

Rad4XCNN: a new agnostic method for post-hoc global explanation of CNN-derived features by means of radiomics

TL;DR

This paper presents a novel method, namely Rad4XCNN, to enhance the predictive power of CNN-derived features with the inherent interpretability of radiomic features, by associating intelligible meaning to CNN-derived features by means of Radiomics, offering new perspectives on explanation methods beyond visualization maps.

Abstract

In recent years, machine learning-based clinical decision support systems (CDSS) have played a key role in the analysis of several medical conditions. Despite their promising capabilities, the lack of transparency in AI models poses significant challenges, particularly in medical contexts where reliability is a mandatory aspect. However, it appears that explainability is inversely proportional to accuracy. For this reason, achieving transparency without compromising predictive accuracy remains a key challenge. This paper presents a novel method, namely Rad4XCNN, to enhance the predictive power of CNN-derived features with the inherent interpretability of radiomic features. Rad4XCNN diverges from conventional methods based on saliency maps, by associating intelligible meaning to CNN-derived features by means of Radiomics, offering new perspectives on explanation methods beyond visualization maps. Using a breast cancer classification task as a case study, we evaluated Rad4XCNN on ultrasound imaging datasets, including an online dataset and two in-house datasets for internal and external validation. Some key results are: i) CNN-derived features guarantee more robust accuracy when compared against ViT-derived and radiomic features; ii) conventional visualization map methods for explanation present several pitfalls; iii) Rad4XCNN does not sacrifice model accuracy for their explainability; iv) Rad4XCNN provides a global explanation enabling the physician to extract global insights and findings. Our method can mitigate some concerns related to the explainability-accuracy trade-off. This study highlighted the importance of proposing new methods for model explanation without affecting their accuracy.
Paper Structure (42 sections, 7 equations, 5 figures, 5 tables)

This paper contains 42 sections, 7 equations, 5 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Explainable AI
  4. Explainability in Deep Architectures
  5. Explainability in Radiomics
  6. Our Rationale
  7. Materials and Methods
  8. Datasets Description
  9. Breast Ultrasound Images Dataset (BUSI) -
  10. Palermo and Cefalù Datasets -
  11. Deep Architecture Training
  12. Image preprocessing -
  13. Training and test protocol -
  14. Saliency Maps Computation -
  15. Radiomic Models Training
  16. ...and 27 more sections

Figures (5)

  • Figure 1: Overall workflow. Deep and Radiomic approaches were implemented for feature extraction and classification. Successively, radiomic and deep features were employed for MethodAcronym implementation, providing a global explanation method.
  • Figure 2: Diagram depicting the details about the training, the testing, and the external validation related to deep and radiomic workflow.
  • Figure 3: (a) original ultrasound image; (b) ultrasound image with overlapped the GradCAM saliency map; (c) ultrasound image with overlapped the EigenCAM saliency map; (d) ultrasound image with overlapped the ScoreCAM saliency map.
  • Figure 4: Global explanation of CNN-derived features obtained using MethodAcronym. Each bar represents the number of CNN-derived features showing a correlation with radiomic features. The Y-axis reports the number of deep features correlated with the radiomic feature specified in the X-axis.
  • Figure 5: Correlation trend between radiomic and deep features.