Analysis of Breast Thermograms Using Asymmetry in Infra-Mammary Curves

Automated tumor segmentation in thermographic breast images

Identifying and delineating suspicious regions in thermal breast images poses significant challenges for radiologists during the examination and interpretation of thermogram images. This paper aims to tackle concerns related to enhancing the differentiation between cancerous regions and the background to achieve uniformity in the intensity of breast cancer's (BC) existence. Furthermore, it aims to effectively segment tumors that exhibit limited contrast with the background and extract relevant features that can distinguish tumors from the surrounding tissue. A new cancer segmentation scheme comprised of two primary stages is proposed to tackle these challenges. In the first stage, an innovative image enhancement technique based on local image enhancement with a hyperbolization function is employed to significantly improve the quality and contrast of breast imagery. This technique enhances the local details and edges of the images while preserving global brightness and contrast. In the second stage, a dedicated algorithm based on an image-dependent weighting strategy is employed to accurately segment tumor regions within the given images. This algorithm assigns different weights to different pixels based on their similarity to the tumor region and uses a thresholding method to separate the tumor from the background. The proposed enhancement and segmentation methods were evaluated using the Database for Mastology Research (DMR-IR). The experimental results demonstrate remarkable performance, with an average segmentation accuracy, sensitivity, and specificity coefficient values of 97%, 80%, and 99%, respectively. These findings convincingly establish the superiority of the proposed method over state-of-the-art techniques. The obtained results demonstrate the potential of the proposed method to aid in the early detection of breast cancer through improved diagnosis and interpretation of thermogram images.

LINPE-BL: A Local Descriptor and Broad Learning for Identification of Abnormal Breast Thermograms

This paper proposes a novel local feature descriptor coined as a local instant-and-center-symmetric neighbor-based pattern of the extrema-images (LINPE) to detect breast abnormalities in thermal breast images. It is a hybrid descriptor that combines two different feature descriptors: one is the inverse-probability difference extrema (IpDE), and another is the local instant and center-symmetric neighbor-based pattern (LICsNP). IpDE is developed to compute the intensity-inhomogeneity-invariant feature-based image of the breast thermogram. Besides, the LICsNP is intended to capture the local microstructure pattern information in the IpDE image. A new paradigm, named Broad Learning (BL) network, is introduced here as a classifier to differentiate the healthy and sick breast thermograms efficiently. The efficacy of the proposed system is quantitatively validated on the images of DMR-IR and DBT-TU-JU databases. Extensive experimentation on these databases with an average accuracy of 96.90% and 94%, respectively, justifies proposed system's superiority in the differentiation of healthy and sick breast thermograms over the other related existing state-of-the-art methods. The proposed system also performs consistently in the presence of noise and rotational changes.

Highly Discriminative Physiological Parameters for Thermal Pattern Classification

Infrared Thermography (IRT) is a non-contact, non-intrusive, and non-ionizing radiation tool used for detecting breast lesions. This paper analyzes the surface temperature distribution (STD) on an optimal Region of Interest (RoI) for extraction of suitable internal heat source parameters. The physiological parameters are estimated through the inverse solution of the bio-heat equation and the STD of suspicious areas related to the hottest spots of the RoI. To reach these values, the STD is analyzed by means: the Depth-Intensity-Radius (D-I-R) measurement model and the fitting method of Lorentz curve. A highly discriminative pattern vector composed of the extracted physiological parameters is proposed to classify normal and abnormal breast thermograms. A well-defined RoI is delimited at a radial distance, determined by the Support Vector Machines (SVM). Nevertheless, this distance is less than or equal to 1.8 cm due to the maximum temperature location close to the boundary image. The methodology is applied to 87 breast thermograms that belong to the Database for Mastology Research with Infrared Image (DMR-IR). This methodology does not apply any image enhancements or normalization of input data. At an optimal position, the three-dimensional scattergrams show a correct separation between normal and abnormal thermograms. In other cases, the feature vectors are highly correlated. According to our experimental results, the proposed pattern vector extracted at optimal position a=1.6 cm reaches the highest sensitivity, specificity, and accuracy. Even more, the proposed technique utilizes a reduced number of physiological parameters to obtain a Correct Rate Classification (CRC) of 100%. The precision assessment confirms the performance superiority of the proposed method compared with other techniques for the breast thermogram classification of the DMR-IR.