[Paper Review] Comparing Results of Thermographic Images Based Diagnosis for Breast Diseases

Érick Oliveira Rodrigues, Aura Conci+4 more|arXiv (Cornell University)|Aug 30, 2022

Infrared Thermography in Medicine8 references18 citations

TL;DR

This study evaluates infrared thermography for breast disease detection using a public dataset of 102 static thermographic images (54 normal, 48 with findings) from UFPE Hospital. It compares machine learning classifiers, achieving 61.7% accuracy and a Youden index of 0.24 using the Sequential Minimal Optimization (SMO) algorithm on features from a top-performing prior method.

ABSTRACT

This paper examines the potential contribution of infrared (IR) imaging in breast diseases detection. It compares obtained results using some algorithms for detection of malignant breast conditions such as Support Vector Machine (SVM) regarding the consistency of different approaches when applied to public data. Moreover, in order to avail the actual IR imaging's capability as a complement on clinical trials and to promote researches using high-resolution IR imaging we deemed the use of a public database revised by confidently trained breast physicians as essential. Only the static acquisition protocol is regarded in our work. We used lO2 IR single breast images from the Pro Engenharia (PROENG) public database (54 normal and 48 with some finding). These images were collected from Universidade Federal de Pernambuco (UFPE) University's Hospital. We employed the same features proposed by the authors of the work that presented the best results and achieved an accuracy of 61.7 % and Youden index of 0.24 using the Sequential Minimal Optimization (SMO) classifier.

Motivation & Objective

• To assess the diagnostic potential of infrared thermography in detecting breast abnormalities using machine learning.
• To compare the performance of different classification algorithms on a standardized public thermographic dataset.
• To validate the utility of high-resolution static thermographic imaging as a complementary tool in clinical breast disease screening.
• To establish a benchmark using features from the best-performing prior method on a revised, physician-verified public database.
• To promote research in thermographic imaging by providing a reliable, publicly accessible dataset with clinical validation.

Proposed method

• Utilized 102 static infrared thermographic images from the Pro Engenharia (PROENG) public database, collected at UFPE University Hospital.
• Applied the same feature set used in the highest-performing prior study to ensure consistency and comparability.
• Employed the Sequential Minimal Optimization (SMO) classifier for binary classification of normal vs. abnormal cases.
• Evaluated model performance using standard metrics including accuracy and the Youden index.
• Focused exclusively on static acquisition protocols, excluding dynamic or thermoregulatory response analyses.
• Used a physician-verified, publicly available dataset to ensure clinical relevance and reproducibility.

Experimental results

Research questions

• RQ1 Can infrared thermography achieve clinically meaningful accuracy in distinguishing malignant breast conditions from normal cases using machine learning?
• RQ2 How does the performance of the SMO classifier compare to other algorithms when applied to a standardized thermographic dataset?
• RQ3 To what extent does using a physician-verified, high-resolution public database improve the reliability of thermographic diagnosis research?
• RQ4 What is the diagnostic utility of static thermographic imaging alone, as measured by accuracy and the Youden index?
• RQ5 Can consistent feature extraction from prior top-performing models be replicated to achieve comparable results on a new dataset?

Key findings

• The SMO classifier achieved a diagnostic accuracy of 61.7% on the test set of 102 thermographic images.
• The Youden index of 0.24 indicates a moderate level of diagnostic discrimination, suggesting limited but non-trivial separation between normal and abnormal cases.
• The study confirms that thermographic imaging, when combined with machine learning, can yield measurable diagnostic performance on public datasets.
• The results demonstrate reproducibility of prior findings using a standardized feature set and a physician-verified dataset.
• The performance level suggests thermography may serve as a complementary tool in clinical screening but is unlikely to replace standard imaging modalities.
• The use of a public, high-resolution, physician-reviewed database enhances the reliability and reproducibility of future thermographic research.