Simultaneous localization of multiple tumors from thermogram of tissue phantom by using a novel optimization algorithm inspired by hunting dogs

Highlights

• Experiments conducted on tissue-mimicking phantoms containing tumors.

• Extract thermograms on the sample surface by an examiner robot.

• Introduce a novel optimization algorithm inspired by the behavior of hunting dogs.

• Diagnosis of cases with multiple tumors and tumors smaller than 0.3 mm in diameter.

• Significant reduction of examination time to only a few seconds by Hunter algorithm.

Abstract

The objective of this study is to couple the contact thermography method with a novel optimization algorithm to rapidly detect and localize the soft tissue tumor. To this end, experiments are carried out on tissue-mimicking phantoms containing resistance heaters to simulate the embedded tumors. An examiner robot is used to measure the temperature of the tissue surface. The time required for the examination of the tissue surface is reduced by developing a novel optimization algorithm called the Hunter Algorithm (HA). In the HA, population individuals are called the hunters, and the global maximum is referred to as the prey. The maximum temperature occurs at the location of the tumor. By the end of the hunting procedure, a flock of hunters converges to the maximum temperature and reaches the tumor while the examination time is significantly reduced. Performance of the HA is evaluated by applying the Genetic Algorithm (GA) and the Particle Swarm Optimization (PSO) algorithm to 11 test functions as minimization problems. It is observed that for the Ackley's function, as an example, the HA finds the global minimum after the 10th iteration with an accuracy of ${10}^{-4}$, while the PSO converges with the same accuracy after the 30th iteration and the accuracy of the GA remains about 0.002. In addition, the results show that the contact thermography in conjunction with the HA is of clinical importance in accurate detection of multiple tumors and small and deeply located tumors with insignificant thermal effects on the tissue surface.

Introduction

Constant heat is generated by the human body and is dissipated by a thermoregulatory system [1]. This phenomenon results in the maintenance of the normal body temperature. Disorders and diseases affect the body normal temperature and may increase it up to 2–3 °C. Thermography is a developing non-invasive technique for the diagnosis of diseases which cause an elevation in the body temperature. Thermography provides a map of the body surface temperature, which is referred to as the thermogram [1]. In recent decades, extensive researches have been conducted on the thermography method to determine a close relationship between the body physiology and the temperature of the tissue surface. The success of this technique is verified in the diagnosis of breast cancer [2,3], eye disease [4,5], skin cancer [6,7], joint disease, and bone and thyroid tumors [8,9]. In the presence of a tumor, hot spots can be seen on the thermogram, which is due to a higher surface temperature around the area of cancerous cells compared to the healthy tissue [10]. Thermograms revealed a clear temperature difference up to 3.3 °C, between the tumor core (36.4 °C) and the surrounding healthy tissue (33.1 °C) [11].

Based on the method used for the temperature measurement, thermography can be categorized as contact and infrared. In the infrared (IR) thermography, the infrared energy emitted from an object is detected by an infrared camera or an infrared sensor, and it is converted to apparent temperature on a thermogram [10]. Many types of research have been focused on the ability of the IR thermography in analyzing the physiological condition of the soft tissue. Francis et al. [12] made use of a rotational thermography system for the diagnosis of breast cancer. It was observed that for the healthy breast, there is a continuous temperature band, which could be disturbed by an abnormality. The main drawback of using the IR thermography in their research and similar researches is neglecting the effect of the surface curvature. Since the temperature value depends on the distance between the IR camera/sensor and the tissue surface, neglecting the surface curvature would result in inaccurate estimations. In the IR thermography, the basis of tumor detection is the uniformity disruption in the thermogram; therefore, the difficulty in maintaining similar distance from the IR camera for all points on the surface is a disadvantage of the IR thermography. Shi et al. [13] used the thermal thermography for the diagnosis of skin tumors. In their study, an estimation methodology was presented to determine unknown thermo-physical parameters of a tumor by using the temperature profile on the skin surface. The temperature profile in the tumoral region was interpolated by the bio-heat transfer equation, and the unknown parameters were estimated. The infrared thermography was used as a possible method for obtaining the thermogram. A disadvantage of using the IR thermography in their research and similar researches is the sensitivity of the IR camera/sensor, which limits the functionality of the IR thermography to the superficial layer of the soft tissue. Therefore, small and deeply located tumors, which produce insignificant thermal effects on the tissue surface, would remain undetected. Bahramian et al. [14] proposed an infrared thermal imaging system for the human neck and the thyroid gland. The thermograms were analyzed to distinguish thyroid tumors from the healthy thyroid tissue. The neck curvature and an inadequate accuracy of the IR camera were the main drawbacks of this study. Commonly, to compensate the lack of measurement accuracy, the IR detector is located closer to the object. Unfortunately, the measurement error usually increases by reducing the distance due to the effects of reflectivity and the inclusion of other heat sources within the sensor's field of view.

Contact thermography is used for cancer detection along with the IR thermography. In this method, the temperature is measured by contact temperature sensors on the tissue surface. This method is inspired by the clinical or the physical examination by a physician. Physical examination is the first screening method for early cancer detection. By using a high precision temperature sensor, the accuracy of the contact thermography could be improved. Therefore, regular use of contact thermography could be considered as an inexpensive screening method, which can prevent cancer progression. Contact thermography has been used by Sadeghi et al. [[15], [16], [17]] in recent years for the detection of brain tumors. They have proposed a thermal imaging system for the intraoperative detection of brain tumors. The system was equipped with a contact temperature sensor with a high accuracy. It has been observed that temperature map on the tissue surface is an appropriate outcome of the contact thermography for superficial tumors. Meanwhile, the heat flux map on the surface could be used as an extra thermal result for deeply located tumors. The thermograms were analyzed to localize the brain tumor and identify the tumor margins. The contact thermography has the disadvantage of slow response compared to the IR thermography. However, due to the advantages of higher accuracy in the range of biological temperature and independence from the surface geometrical shape, the contact thermography is preferred and selected in the present study.

In the present study, the contact thermography is employed to detect soft tissue tumor, while the method performance is improved by using the artificial tactile sensing (ATS) method. In the ATS, the tissue surface is slightly compressed during the tissue examination. It has been already affirmed that the tissue compression can improve the quality of thermal measurements, significantly [15]. Therefore, before the temperature measurement, we apply a small compressive strain on the tissue surface. Afterward, the thermogram is constructed by measuring the temperature of different test points on the tissue surface. Tumor detection and localization are carried out by analyzing the thermograms. For accurate tumor detection, several test points should be examined on the tissue surface, which increases the examination time. To overcome this problem, a novel optimization algorithm is introduced to reduce the number of temperature measurements on the tissue surface and to rapidly detect and localize the tumor.

Optimization techniques are extensively used to solve different optimization problems in applied sciences, engineering, industry, biology, and computer science. Some well-known optimization algorithms such as Genetic Algorithm (GA) [18], Particle Swarm Optimization (PSO) [19], Bee Colony Optimization (BCO) [20], Ant Colony Optimization (ACO) [21], Simulated Annealing (SA) [22] and hunting search algorithm [23] are used in this regard. However, these algorithms suffer from significant deficiencies such as the requirement to a large number of input parameters, being unpredictable, producing unripe results and lots of iterations for convergence achievement [24]. Therefore, looking for high-efficiency algorithms is an essential problem in the rapid contact thermography application.

In this regard, a new optimization algorithm is proposed, which is called the Hunter Algorithm (HA) and is inspired by the hunting dogs' behavior. In the HA, all the hunters do their best to hunt the prey on the hunting ground. Following the smell of the prey, leads the hunters to converge to the global maximum on the hunting ground. In this study, the hunting ground, the prey, and the smell are identical to the thermogram, tumor, and the temperature. In addition to improving the rate of convergence, the novelty of the HA is the requirement to only one hunter as the initial population, to find the prey. This is of particular importance in the tumor detection procedure in the contact thermography. In practice, use of one temperature sensor instead of multiple sensors (as the hunters) for scanning the tissue surface would prevent the crosstalk effect on the measured values.

In section 2.2, an experimental setup is described, which mimics the soft tissue containing a tumor. Section 2.3 to section 2.7 explain how the hunting procedure is modeled and implemented among the hunters on the hunting ground. In section 3, the result of rapid tumor detection by the hunting algorithm is provided.