Skip to main content

A review of computer assisted detection/diagnosis (CAD) in breast thermography for breast cancer detection

May 7, 2021

Mehrdad Moghbel · Syamsiah Mashohor

Abstract  Breast cancer is the leading type of cancer diagnosed in women. For years human limitations in interpreting the thermograms possessed a considerable challenge, but with the introduction of computer assisted detection/diagnosis (CAD), this problem has been addressed. This review paper compares different approaches based on neural networks and fuzzy systems which have been implemented in different CAD designs. The greatest improvement in CAD systems was achieved with a combination of fuzzy logic and artificial neural networks in the form of FALCON-AART complementary learning fuzzy neural net- work (CLFNN). With a CAD design based on FALCON-AART, it was possible to achieve an overall accuracy of near 90%. This confirms that CAD systems are indeed a valuable addition to the efforts for the diagnosis of breast cancer. Lower cost and high performance of new infrared systems combined with accurate CAD designs can promote the use of thermography in many breast cancer centres worldwide.

Keywords Thermography · Breast cancer · Computer aided detection/diagnosis · Fuzzy logic · Neural networks

1 Introduction

In 1956 Lawson suggested that the skin temperature over a cancerous area of the breast and the Venus blood draining from the cancerous tumor were higher than the surround- ing normal tissue (Lawson 1956). In 1965, infrared imaging was introduced to the United States by Gershen-Cohen et al. (1965), from the Albert Einstein Medical Center. Unlike mammography, infrared imaging does not use any ionizing radiation, venous access or any invasive procedures. Therefore, the examination poses no known harm to the patient because thermography is based on the infrared radiation emitted from the human body. All objects with temperature higher than absolute zero emit infrared radiation. The imaging procedure is also simple. The patient is seated in a room with a temperature of 1822 degrees centigrade for a time of 5–15 minutes so an equilibrium between skin temperature and environmental temperature is reached. Then an infrared camera records the infrared radiation emitted from the patient’s body.

This ease of imaging and absence of the problems related to the ionizing radiation associated with mammography brought a high initial interest in Thermography and in the early studies done by the pioneers of this field consisting of more than 60,000 patients (Gershen-Cohen et al. 1965; Hoffman et al. 1967; Stark and Way 1974; ?), thermogra- phy achieved an average specificity and sensitivity of over 88% which were higher than mammography at the time.

This high interest in thermography convinced the NCI (National Cancer Institute) to include thermography in their large scale study to compare different breast cancer detec- tion methods which took place from 1973 till 1979, known as the Breast Cancer Detection and Demonstration Project (BCDDP) (Baker 1982). But due to inconsistent image interpret- ing protocols and poor study design, thermography was dropped from BCDDP. This led to a decreased interest on thermography especially in the US and only a handful of centers contin- ued research and publication on thermography. On the other hand, other countries considered thermography to be a valuable assistance and were more active on the research. Today, many countries are considering thermography as a first line detection system such as japan with more than 1,500 hospitals and clinics which are using thermography as a first-line detection, Korea uses more than 450 centers and there are also many centers currently operating in Germany, Austria, United Kingdom, Poland, Italy, China, United States, Canada, Russia, Norway, Australia and South America (?).

With the declassification of the military infrared sensing technology in the 1990s a new type of uncooled infrared camera hit the market also known as a second generation infrared camera, which almost fixed all the problems faced by the previous generation infrared cam- eras such as thermal drift, poor sensitivity, time required for image acquisition (Head et al. 1996; Anbar 1998).

The availability of higher resolution images coupled with the emphasis on the early detec- tion of cancer brought back the interest in thermography. Thermography has some unique properties such as the ability to detect early warning signs of cancer up to 10 years before the appearance of cancer on other imaging modalities (Gamagami et al. 1997; Head et al. 2000) or its ability to assess the treatment efficiency (Gamagami et al. 1997; Keyserlingk et al. 2001) and the prognosis condition of the patient (Head et al. 1993, 2000; Ohsumi et al. 2002). It was shown that 44% of the patients with abnormal thermograms will develop breast cancer within 5years, also hot cancers showed a significantly poor prognosis with a 24% survival rate at 3 years while cooler cancers showed up to 80% survival rate at 3 years (?).

The sensitivity of mammography decreases on younger persons or women with dense breasts but thermography does not depend on the age of the patient or the density of breasts (Head et al. 1993; Qi and Diakides 2003), In 70% of cases signs of breast cancer will be detectable by infrared imaging up to 1 year before it is diagnosable with mammography (Head et al. 2000). Average tumor size undetected by thermal imaging is 1.28 cm while the average tumor size undetected by mammography is 1.66 cm (Keyserlingk et al. 2002). Thermography has obtained an average sensitivity and specificity of 90% (Head et al. 1999).

As a future risk indicator for breast cancer a persistent abnormal thermograph carries a 22 time higher risk and is 10 times more significant than a first order family history of the disease (Head et al. 2002; ?). In recent years with the advances in image processing techniques and Computer Assisted Detection (CAD) thermography has achieved an average sensitivity and specify higher than mammography (Ng et al. 2005; Schaefer et al. 2007; Tan et al. 2007; Schaefer et al. 2008).

2 Acceptance of CAD in medical imaging

Early studies on using computers for analyzing medical images date back to 1960s, during that period many believed that computers can replace the physicians and so, much faith was put in computer diagnosis. However, computers were at early stages of development and lacked the processing power, also the advanced image processing techniques which exist today were not available during that time and digital images were rare. This yield to failure of computers in detecting abnormalities with an acceptable performance and the idea was pushed aside. However in 1980s scientists begin using computers in aiding the physicians by highlighting the problematic areas or helping physicians by providing second opinion on the diagnosis of abnormalities. This concept known as Computer Assisted Detection/Diagnosis (CAD) became widely accepted around the world. CAD does not try to replace the physi- cian but rather works on the principle of aiding the physician on the diagnosis by providing additional information on image abnormalities or indicating the parts of image which looks normal to the naked eye but in fact abnormalities exists on that region or by giving a diagnosis result which can serve as a second opinion in decision making. CAD is currently employed in many aspects of daily clinical practice such as breast cancer, skeletal, brain and vascular disorders.

3 Definition of sensitivity, specificity and accuracy in medical imaging

3.1 Definitions of true and false positive/negative

  • True positive: Sick people correctly diagnosed as sick (TP)
  • False positive: Healthy people incorrectly identified as sick (FP)
  • True negative: Healthy people correctly identified as healthy (TN)
  • False negative: Sick people incorrectly identified as healthy (FN)