Breast Cancer Detection Using Infrared Thermography: A Survey of Texture Analysis and Machine Learning Approaches

Abstract

Breast cancer remains a leading cause of cancer-related deaths among women worldwide, highlighting the urgent need for early detection. While mammography is the gold standard, it faces cost and accessibility barriers in resource-limited areas. Infrared thermography is a promising cost-effective, non-invasive, painless, and radiation-free alternative that detects tumors by measuring their thermal signatures through thermal infrared radiation. However, challenges persist, including limited clinical validation, lack of Food and Drug Administration (FDA) approval as a primary screening tool, physiological variations among individuals, differing interpretation standards, and a shortage of specialized radiologists. This survey uniquely focuses on integrating texture analysis and machine learning within infrared thermography for breast cancer detection, addressing the existing literature gaps, and noting that this approach achieves high-ranking results. It comprehensively reviews the entire processing pipeline, from image preprocessing and feature extraction to classification and performance assessment. The survey critically analyzes the current limitations, including over-reliance on limited datasets like DMR-IR. By exploring recent advancements, this work aims to reduce radiologists' workload, enhance diagnostic accuracy, and identify key future research directions in this evolving field.

Keywords: breast cancer; image processing; medical image analysis; texture; thermography.

Figure 1

Right oblique, frontal, and left oblique images of patient IIR0035 from the Mendeley dataset [21]. The red dotted circle shows the thermal pattern of a cancerous lesion in the patient’s right breast. The patient’s left breast is normal. Note that there is a brighter more elaborate pattern in the patient’s right breast.

Figure 2

End-to-end process for a CAD system for breast cancer.

Figure 3

Combining hand-crafted textural with ANN-based feature extraction.

Figure 4

Images showing right lateral, right oblique, frontal, left oblique, and left lateral perspectives from the DMR-IR dataset [39]. The top row is images of patient #90, who is classified as healthy, and the bottom row is images from patient #180 who is classified as abnormal.

Figure 5

Segmentation approaches on patient #180 (abnormal) from DMR-IR dataset [39]: (a–c)—outline front, right, and left breast; (d)—pre-segmented right breast from DMR-IR dataset; (e)—outline breasts in same image; (f,g)—box right and left breasts.

Figure 6

Neighborhood pixels of distance one from center pixel. Pixels 1 and 5 are at an angle $0^{°}$, pixels 4 and 8 at an angle ${45}^{°}$, pixels 7 and 8 at an angle ${90}^{°}$, and pixels 6 and 2 at an angle ${135}^{°}$.

Figure 7

Example calculation of Haralick et al. [78] texture features with $\theta$ = 0 and distance = 1.

Figure 8

LBP operator on a 9 × 9 matrix. The center intensity value 5 is replaced by the intensity value 174.

Figure 9

Comparison of a grayscale image of patient #180 (abnormal) from the DMR-IR dataset [39] with binary encoded CT and LBP setting R = 2 and P = 8.

Figure 10

Applying a Gabor filter bank to a thermographic image of patient #180 (abnormal) from the DMR-IR dataset [39] with phases $0^{°}, {45}^{°}, and {90}^{°}$ and pixel widths of 4 and 8.

Figure 11

Frontal image of patient #180 (abnormal) from the DMR-IR dataset [39]. The explosion of the breast areas shows 8 × 8 cells as green boxes and the regions as red boxes of size 2 × 2 cells, which overlap adjacent regions by one cell. The normalized histogram of gradients for the area is provided as well.