AI Talos: Breast Cancer Detection using Thermal Imaging and Artificial Intelligence

13 min readMar 22, 2021

Abstract

Breast cancer is the most diagnosed cancer among women. There are more than 18 types of breast cancer. It is estimated that over 42,170 people died of breast cancer in the year 2020 and 276,480 new cases were found. Several standard screening techniques are used for early detection of breast cancer such as Mammography and Thermography. In recent years, due to the advancements in computers, infrared cameras, and thermography devices, thermography has been able to achieve more Sensitivity and Specificity than other methods such as mammography. With the production and development of neural networks and also introducing systems based on fuzzy logic as well as the high quality of the Thermograms as a result of the second generation of cameras, Thermography systems (CAD) draw the attention of many researchers. In this study, (i) we will cover the recent research related to the implementation of deep learning methods in the detection of breast cancer using thermography cameras, and (ii) review AI Talos software which is an AI-guided decision-making software for breast cancer screening to assist physicians in early detection of breast cancer, and also (iii) analyze up to 10 random sick/healthy patient’s case studies to evaluate the software’s performance.

Keywords: Breast Cancer; Thermography; Deep Learning; Machine Learning; Computer-Aided Diagnosis; Artificial intelligence

Introduction

Breast cancer is the second leading cause of death among women. Breast cancer can be treated successfully if detected early. However, despite the current progress in medicine, it remained the most common cause of death worldwide due to not being detected early. Since the early diagnosis of breast cancer is very difficult, statistical methods and artificial intelligence (AI) techniques can be very important in this regard. Different screening tools such as mammography, ultrasound, and thermography have been advanced for the early detection of breast cancer. In this way, computers assist doctors or radiologists in identifying abnormalities using image processing techniques and artificial intelligence (AI).

In retrospect

In the 1960s, there was a strong tendency for replacing computers with physicians. However, at that time computers were not developed yet and modern digital images didn’t exist. These were the problems that caused the computer’s failure for detecting the abnormalities.

In the 1980s, despite the failure in the 1960s, scientists used this approach to assist physicians to identify abnormal areas and also provide a second opinion besides doctor’s detection. The approach is called Computer Assisted Diagnosis (CAD) and it is well-accepted all around the World.

Despite the 1960s, This approach (CAD) didn’t attempt to replace physicians with computers but it helped them to reach a more reliable diagnosis. From 1977 to 2015, Several outstanding CAD systems in Thermography have been made. Even in 2002, with images taken from the camera of the second generation and the use of recurrent neural networks, the percentage of accuracy was the biggest challenge. In recent years, due to the advancements in computers, infrared cameras, and thermography devices, thermography has been able to achieve more Sensitivity and Specificity than other methods such as mammography. With the production and development of neural networks and also introducing systems based on fuzzy logic as well as the high quality of the Thermograms as a result of the second generation of cameras, Thermography systems (CAD) draw the attention of many researchers.

Artificial Intelligence in Breast Cancer Thermography

The AI is said to be an artificially intelligent machine in various situations. In other words, these are systems that can respond to similar conditions such as an intelligent human, including understanding complex situations, simulating thinking processes and human reasoning methods, and demonstrating accurate responding, learning and ability to acquire knowledge, and reasoning for solving problems. (DL & AK, 2010, #)

By using image processing pattern recognition, and AI, researchers have been able to provide techniques that accurately detect masses. (Quintanilla-Dominguez J et al., 2010, #)

This article presents AI Talos which is an AI-powered software that uses medical thermography images to detect breast cancer. The findings are presented in the form of different case studies scrutinized by this software and the results are presented in this study to evaluate the software’s AI performance.

AI Talos: Actionable and trustful indication of suspicious cases

At your side and extremely accurate, the uniquely actionable AI Talos software is AI-guided decision-making support for breast cancer thermography screening. This application provides quick and reliable confirmation of what you suspect and what needs more attention. This software gives you increased certainty in your diagnosis. Each image demands deep, informed concentration. This AI software works alongside your viewer to give you additional confidence and provides a high level of Interpretation.

Technology

The technology behind AI Talos software is a low-cost, precise, and innovative medical software that can be used in hospitals, clinics, and any other thermography center. It is a machine automated detecting software that works for women of all ages and can detect breast cancer years before any other screening method. Some of the unique features of its technology are:

Machine Learning on validation of thermal images
Deep Learning, in general, is the core of the artificial intelligence method
Transfer Learning, Segmentation, Feature Learning, …
Cloud processing
High accuracy and best in sensitivity and specificity
Privacy sensitive

Major benefits of AI Talos software are listed in the following paragraphs:

The accurate, actionable interface that boosts your confidence, helping you achieve the highest standard of care
It’s the second set of supervision that adds a safety net to your process
Web panel easily fits into your existing infrastructure, with at-a-glance access to results

Some of the features of AI Talos are also mentioned here:

Web panel that works on all kind of devices (smartphone/desktop)
Automated report generation with unique parameters generated by AI Talos
Artificial intelligence to provide at-a-glance level-of-suspicion scoring of studies; confirming findings that are certain, and questionable cases that require further scrutiny
Automated marking of potential abnormalities in breast thermal images
Simple to use EMR to record patient data & every visits all in one place
Tested by experienced Thermographers around the world

AI Talos is an AI-based software that works with your thermal camera to provide an expert, confirming look at thermal images. Its non-intrusive design does not interrupt the current workflow, instead provides results on a cloud-based web panel. The web-based panel presents all the extra tools you need and synchronizes with your worklist, at once, with no additional software needed. Get access to the AI Talos web panel to upload and manage your patients and their visits for breast cancer screening: https://panel.aitalos.com/login

Enhance productivity, reduce costs and save time

AI Talos provides you with an intuitive, intelligent and integrated breast screening app. After signing up/logging into AI Talos you will have your easy-to-use panel which includes everything you need.

You can store your patient’s records and visits over time for fast searching and careful overviewing. You can simply add new patients to your panel. Patients can be categorized as low-risk, medium-risk, and high-risk patients based on the results derived from Artificial intelligence.

If you are using AI Talos in more than one department or clinic, you can simply add new centers to your panel. They will be added to the panel by their record ID, name, and the name of the owner of the center. AI Talos will connect all your departments in only one panel where you can have access to them by clicking on the search button.

In a medical environment, more than one person needs to be involved when making decisions. AI Talos lets you give access to other members of your medical team using a shared panel. In the user menu, you can simply add new users to your panel where other members of your team can be added. Meanwhile, each person may have a different role. You can assign multiple roles such as owner, operator, thermographer, etc. to the members in the user menu, role list.

In the patient menu, you will have access to the records of your added patients searching them by using different filters such as their record ID, their name, or among different centers. You can also add new patients or visits to the patient menu.

AI Talos is an AI-based software that works as second sets of eyes for you to give you a superpower on finding abnormality & high-risk cases.

Methods & Materials

This work reviews current progress in breast cancer detection using Artificial intelligence and thermography as a non-invasive approach with AI Talos AI-Powered software. The state-of-the-art and contributions of this study can be summarized as follows:

1 — Method & Materials

2 — Trial and results

3 — Conclusion

In this case study, we are using two different databases including healthy and sick patients. Patients’ cases are different in age and methods.

Camera used in the project FLIR SC-620
Resolution 640 x 480: Pixel = 45 μm
78 Sick cases
216 Healthy cases
Race: White, Indian, Black
Age: Various

Method

By using the AI Talos cloud-base panel we’ve uploaded and stored patients’ records in the EHR system, Since we are using an account to proceed with the study we will be able to access this information for future study and evaluation.

The technology behind the AI Talos:

Convolutional neural network
The neural network is inspired by how neurons in the human brain work. Each neuron in the human brain is interconnected and information flows across each of these neurons. In NNs, each neuron receives input and performs a dot operation with weights and biases. Weight specifies the strength of the connection between two nodes. Biases are external values that increase or decrease the net input of the activation function. Nodes are the individual processing units in each layer. Figure 1 illustrates the mathematical model of a neuron.
The activation function expresses a linear combination of input x
Concerning neurons and parameters, followed by an element-wise non-linearity. The transferor activation function determines whether or not the neuron is “active” based on the weighted sum of the input.
Learning from data has two main goals: to understand the data generation process and data interpretation; and to predict future observations. Predicting future projects does not require a probabilistic accuracy rate. However, accuracy is a focus of medical data interpretation. As applied in detecting breast cancer, 100% accuracy is required to ensure that the diagnosis follows the ground truth.
A neural network is a massively parallel distributed processor made up of simple processing units, which has a natural propensity for strong experiential knowledge, making it available for use. The NN algorithm allows the learning of the qualitative value of an image. Thus, it is appropriate for application in breast thermogram classification.
A convolutional neural network (CNN) is a deep neural network algorithm that processes input images by assigning certain learnable weights and biases to map important features that differentiate one image from others. In this way, the classification result can be observed as the output. Figure 5 shows the general architecture of CNNs for classifying the breast thermograms into two classes, healthy and cancer. Three major considerations must be made: dataset preparation in image preprocessing, feature learning, and classification. The classification can be binary (healthy and cancer), or more classes such as healthy, benign, and malignant. In the following sections, we review the concepts and related efforts in CNN implementation for breast thermogram classification.

Trial & Result

For this case study, we have chosen random cases from Visual Lab DMR — Database For Mastology Research then we labeled them by their status, So the name of patient’s whose status confirmed to be cancerous is SICK and we added numbers to differentiate them from each other. The same procedure is done for HEALTHY patients.

Case#1 — Sick001

Record: 488749 in DMR database

Patient, 89 years old

Registered at 2013–08–27 (y-m-d)

Screenshot of AI Talos report — Result: Risk Score: 8.48 — High Risk

Case#2 — Sick002

Record: 386049 in DMR database

Patient, 59 years old

Registered at 2013–11–05 (y-m-d)

Screenshot of AI Talos report — Result: Risk Score: 7.60 — High Risk

Case#3 — Sick003

Record: 284553 in DMR database

Patient, 62 years old

Registered at 2013–11–12 (y-m-d)

Screenshot of AI Talos report — Result: Risk Score: 9.71 — High Risk

Case#4 — Sick004

Record: 465813 in DMR database

Patient, 51 years old

Registered at 2014–09–08 (y-m-d)

Screenshot of AI Talos report — Result: Risk Score: 9.40 — High Risk

Case#5 — Sick005

Record: 279981 in DMR database

Patient, 63 years old

Registered at 2014–11–12 (y-m-d)

Screenshot of AI Talos report — Result: Risk Score: 9.56 — High Risk

Case#6 — Healthy001

Volunteer, 65 years old

Registered at 2012–10–30 (y-m-d)

Screenshot of AI Talos report — Result: Risk Score: 2.64 — Low Risk

Case#7 — Healthy002

Volunteer, 59 years old

Registered at 2012–10–31 (y-m-d)

Screenshot of AI Talos report — Result: Risk Score: 3.67 — Low Risk

Case#8 — Healthy003

Record: 663706 in DMR database

Patient, 71 years old

Registered at 2012–10–31 (y-m-d)

Screenshot of AI Talos report — Result: Risk Score: 1.45 — Low Risk

Case#9 — Healthy004

Record: 693388 in DMR database

Patient, 60 years old

Registered at 2012–11–07 (y-m-d)

Screenshot of AI Talos report — Result: Risk Score: 2.14 — Low Risk

Case#10 — Healthy005

Volunteer, 43 years old

Registered at 2012–11–07 (y-m-d)

Screenshot of AI Talos report — Result: Risk Score: 2.14 — Low Risk

To make it clear how this process works we did capture the screening video from the AI Talos panel: https://youtu.be/BtwNsxKEu88

We have repeated the procedure for 100 cases and the result was %95.53 accuracies overall and for sick cases: %93.33 and for healthy cases: %93.33 accuracies.

This study has been filed in the AI Talos cloud base panel which can be accessed by an account. We are willing to grant permission for researchers, clinics, and students under the NDA contract.

For this matter please contact us at research@AiTalos.com

TABLE OF STUDY’S RESULT

1 to 10.99 (from low risk to high risk)

Diagnosis

Cases

AI Talos Risk Measurement Score

Accuracy

Sick

8.822727273

97%

Healthy

3.672666667

96%

Cases risk scores

Conclusion

Today, one of the most common diseases that play a leading role in the death of women is breast cancer. To estimate how breast cancer can develop within a year or so is very difficult. There are different types of breast cancer that can grow at a variety of rates and several factors can have a potential impact on its growth and chance of developing. According to several studies, by using highly well-aimed and accurate soft-wares, breast cancer can be detected and treated at the early stages of the disease. Several applications have been found in the medical profession for thermal imaging. The asymmetrical thermal distribution on breast thermograms can be evaluated using computer-assisted technology like AI Talos. The use of this technology ( AI Talos ) can minimize errors. By using AI Talos AI-Powered software a risk-free screening method using thermography could then be proposed for the self-breast screening method at an early stage without requiring physical involvement. (Manny Movahedi, n.d., #)

Bibliography

Breast Thermography Journal. (n.d.). The Breast Thermography Journal. http://advancedbreastthermography.blogspot.com/2014/

DL, P., & AK, M. (2010). Artificial Intelligence: Foundations of Computational Agents. New York: Cambridge University Press.

Manny Movahedi. (n.d.). AI Talos: A case study of AI-Powered software for breast cancer screening through thermography using 95 patients’ thermal images. https://www.researchgate.net/publication/349099039_AI_Talos_A_case_study_of_AI-Powered_software_for_breast_cancer_screening_through_thermography_using_95_patients_thermal_images

Quintanilla-Dominguez J, Cortina-Januchs MG, Ojeda-magaña B, Jevtić A, Vega-Corona A, & Andina D. (2010). Microcalcification detection applying artificial neural networks and mathematical morphology in digital mammograms. World Automation Congress. New Jersey: IEEE.