Introduction

One of the most deathly type of cancers - the breast cancer, is one of the most serious concerns in modern century and one of the toughest challenges for healthcare and medicine. The scope of the problem currently looks like this: 

• According to WHO in 2020 2.3 million women were diagnosed with breast cancer, and at the end of the year 7.8 women alive were diagnosed with the disease in the past 5 years, making it the world’s most prevalent cancer. 
• Mammography is the main test used for diagnosing breast cancer. Nevertheless, this method has some constraints. For example, mammography is used only on people aged 40+ as very often it is not applicable to young persons. The difficulty comes from the fact that the young breasts are made mainly from glandular tissue which makes them more dense. The high density of the breast interferes with the identification of masses and micro-calcifications by X-rays. This method is uncomfortable and can be painful,  as there is a strong pressing of the breast against the mammogram. For younger women often breast ultrasound is used. But this approach may miss some small lumps or solid tumors that mammography can detect. 
• Early detection of the disease is essential since the chance of curing breast caner drops considerably if not discovered on time.  That’s why, earlier the condition is diagnosed the more effective treatment can be and respectively, the mortality rates may potentially decrease. Statistics show that the 5-year relative survival rate for women in the United States with non-metastatic invasive breast cancer is 91%.

Infrared Thermography (IRT), AI and Breast Cancer

Except for mammography and ultrasound, other diagnosing methods used are tomosynthesis, magnetic resonance imaging and computed thermography. However, all these ways have some downgrades. For example mammography can be extremely painful, and the other methods may not be that precise and require the use of contrast.  

Here is where thermography steps as an effective diagnosing method. It is a low-cost technique that does not involve harmful radiation to humans and it is non-invasive, as well. It is based on the principle of measuring the infrared radiation emitted by an object or surface, through an infrared camera for example, to determine its temperature.

Recent studies combine thermography with AI for achieving the best results in detecting breast cancer because this can provide early diagnoses and thus, higher chances of surviving the disease. Here the application of neural networks in breast cancer detection has a major advantage over traditional methods in terms of time taken for examination because they examine a large amount of data for a short time and predict outputs with high accuracy.

Experiments conducted with IRT and CNN  

Literature in that direction proves the relevance of IRT and AI, and most specifically Convolution Neural Networks (CNN) is a neural network class mostly employed to examine, identify or classify images as it simplifies the images for better analysis. This network is advantageous as it needs fewer human efforts and pre-processing. Back propagation is also included in the learning process to make the network more accurate. Some of the studies in this direction state the following: 

• Dabeer et. el introduces and assesses a deep learning architecture for automated breast cancer detection that incorporates concepts of machine learning and image classification. They described different Deep Neural Networks architectures, especially those adapted to image data such as CNN. This used the labeled (benign/malignant) input image from the raw pixels and highlighted the visual patterns, and then utilize those patterns to distinguish between non-cancerous and cancer containing tissue, working akin to digital staining, which spotlights image segments crucial for diagnostic decisions, with the help of a classifier network. 7009 images have been used in this approach and the accuracy of data is proven to be 99.86%.
• Bardou et. el compare two machine learning approaches for classification of breast cancer detection histology images into benign and malignant subclasses. The first method is trained by support vector machines and it is based on the extraction of a set of handcrafted features encoded by two coding models  while the second is based on the design of CNN. The study showed convolutional neural networks outperformed the handcrafted feature based classifier, where the achieved accuracy is between 96.15% and 98.33% for the binary classification and 83.31% and 88.23% for the multi-class classification.
• Nrea et el proposed CNN for breast mass detection. The model  detects mass region and classifies them into benign or malignant abnormality in mammogram(MG) images, collected from different hospitals, at once The images were passed through different preprocessing stages such as gaussian filter, median filter, bilateral filters and extracted the region of the breast from the background of the MG image. The performance of the model on the test dataset is found to be: detection accuracy 91.86% and sensitivity of 94.67%.

Conclusion

Experiments about cancer identification using thermographic images and classifying them with neural networks show promising results about the application of technology in healthcare. The main objective is for breast cancer to be identified in earlier stages and thus, the mortality rate to be reduced. The earlier the prognosis the better the chance of cure. What’s more, it is a painless, non-invasive process, the cost for equipment is low and the data is analyzed within a matter of second which saves precious time for both patients and specialists, as well. 

‍

References: 

Caroline B Gonçalves, Amanda C. Q. Leles, Lucimara E. Oliveira 2 , Gilmar Guimaraes , Juliano R. Cunha 3 and Henrique Fernandes. Machine Learning and Infrared Thermography for Breast Cancer Detection. Presented at the 15th International Workshop on Advanced Infrared Technology and Applications (AITA 2019), Florence, Italy, 17–19 September 2019. https://www.mdpi.com/2504-3900/27/1/45 

Freer, P.E. Mammographic Breast Density: Impact on Breast Cancer Risk and Implications for Screening. RadioGraphics 2015, 35, 302–315. doi:10.1148/rg.352140106.

Lessa, V.; Marengoni, M. Applying Artificial Neural Network for the Classification of Breast Cancer Using Infrared Thermographic Images. In International Conference on Computer Vision and Graphics; Springer: Berlin, Germany, 2016; pp. 429–438.  

Lessa, V.; Marengoni, M. Applying Artificial Neural Network for the Classification of Breast Cancer Using Infrared Thermographic Images. In International Conference on Computer Vision and Graphics; Springer: Berlin, Germany, 2016; pp. 429–438.

Kandlikar, S.G.; Perez-Raya, I.; Raghupathi, P.A.; Gonzalez-Hernandez, J.L.; Dabydeen, D.; Medeiros, L.; Phatak, P. Infrared imaging technology for breast cancer detection–Current status, protocols and new directions. Int. J. Heat Mass Transf. 2017, 108, 2303–2320 

S. Dabeer, M.M. Khan, S. Islam. Cancer diagnosis in histopathological image: CNN based approach. Inf Med Unlocked, 16 (2019), p. 100231, 10.1016/j.imu.2019.100231

D. Bardou, K. Zhang, S.M. Ahmad. Classification of Breast Cancer Based on Histology Images Using Convolutional Neural Networks. IEEE Access, 6 (2018), pp. 24680-24693

S.H. Nrea, Y.G. Gezahegn, A.S. Boltena, G. Hagos. Breast Cancer Detection Using Convolutional Neural Networks. AI4AH ICLR, 2020 (2020), pp. 1-8

‍