Method for estimating average grey-level's measurement uncertainty from ultrasound images for non-invasive estimation of temperature in different tissue types

Highlights

• Ultrasound is a feasible tool for noninvasive thermometry.

• To estimate measurement uncertainty in US noninvasive thermometry stills a challenge.

• Average grey-levels calculated from B-Mode images can estimate temperature variations.

• Assessing AVGL for different tissues in the same sample leads to uncertainty of 0.68 °C.

Abstract

The objective of this work is to assess, on metrological basis, the average grey-levels (AVGL) calculated from B-Mode images for estimating temperature variations non-invasively in different kinds of tissues. Thermal medicine includes several thermal therapies, being hyperthermia the most noted and well known. Recently, efforts have been made to understand the benefits of ultrasound hyperthermia at mild temperature levels, i.e., between 39 °C and 41 °C. Moreover, the best practices on ultrasound bio-effects research have been encouraged by recommending that temperature rise in the region of interest should be measured even when a thermal mechanism is not being tested. In this work, the average grey-levels (AVGL) calculated from B-Mode images were assessed for non-invasive temperature estimation in a porcine tissue sample containing two different tissue types, fat and muscle, with temperature varying from 35 °C to 41 °C. The sample was continuously imaged with an ultrasound scanner, and simultaneously the temperature was measured. The achieved results were assessed under the light of the measurement uncertainty in order to allow comparability among different ultrasound thermometry methods. The highest expanded uncertainty of estimating temperature variation using AVGL was determined as 0.68 °C.

Introduction

The purpose of this work is to assess, on metrological basis, the average grey-levels (AVGL) calculated from B-Mode images for estimating temperature variations non-invasively in different kinds of tissues. Thermal medicine encloses techniques that manipulate tissue temperature for the treatment of disease, and is being applied since ancient times [1]. Those techniques include several thermal therapies, being hyperthermia the most noted and well known. Hyperthermia is directed to cancer treatment in combination with radiotherapy and chemotherapy. Early studies in hyperthermia considered temperature at cytotoxic levels (42 °C to 45 °C) at the tumour mass, resulting in considerable tumour cell killing. However, difficulties in achieving stable temperatures at these levels and accurate monitoring were reported, especially for irregular and heterogeneous tumours [2]. More recently, efforts have been made to understand the benefits of hyperthermia at mild temperature levels, i.e., between 39 °C and 41 °C. These levels are easy to obtain and both preclinical and clinical data results have demonstrated improved antitumour immune responses with the addition of mild hyperthermia [2]. Another thermal therapy target to cancer is called heat-activated drug delivery, and aim to use thermosensitive nanoparticle drugs carriers that releases chemotherapy agents when heated above 40 °C. The combination of localized hyperthermia with this drug delivery process enables targeted drug delivery.

The Safety Committee of The British Medical Ultrasound Society encourages the best practices on ultrasound bio-effects research, recommending that temperature rise in the region of interest should be measured even when a thermal mechanism is not being tested [3]. Minimally invasive or completely non-invasive methods are desirable to assess temperature, and several methods have been proposed to achieve this goal [4]. Magnetic Resonance Imaging (MRI) has been accepted as the technology able to reach suitable results [5]. However, the high cost, geometrical and technological restrictions of MRI equipment has motivated researchers to explore temperature estimation methods based on other techniques. Ultrasound (US) has been studied as an alternative, considering its low cost and application versatility [6], [7]. The main ultrasound-based methods include echo-signals analysis, where attributes such as changes in time of flight [6], [7], temporal echo-shifts [8], [9], [10], [11], frequency-shifts [11], [12], [13], phase changes due to temporal echo-shifts [11], [14], and changes on the backscattered energy (CBE) [15], [16], [17], [18], [19], [20] are correlated with temperature changes. The CBE has shown a nearly monotonic relation within the hyperthermia temperature range [15], [21], [22]. Bazán et al. (2009) [11] studied the first three methods and concluded that they enable to establish, with a high degree of confidence, linear relations with temperature. However, all of them present advantages and limitations based on three criteria: computation time; temperature resolution, and robustness. For example, time domain estimation may present a good robustness, but require high sampling rates (of at least 1 GS/s). On the other hand, 40 MS/s is sufficient to achieve appropriate results for the method based on frequency-shifts. The phase estimation presents a better resolution than time estimation, but the phase estimation requires signal pre-processing.

Literature also reports changes on the grey-level content of B-Mode images caused by temperature-related changes on backscattered energy [23], [24], [25]. The fundamental physical assumption is that temperature variations induce wave propagation changes that modify the backscattered ultrasound signal having an expression in ultrasonography images. The main effects are apparent motion of the image produced by changes in speed of sound, and changes in image intensity caused by alterations in the amplitude of the echo signal, related with variations in the attenuation and backscattered properties. Teixeira et al. (2014) [26] have shown that the average grey-level (AVGL) values calculated from B-mode images can be used to track temperature changes, while a bovine muscle sample was heated by means of physiotherapeutic ultrasound at 1 MHz and 2.0 W/cm².

In [26], the authors demonstrate that AVGL technique has the ability to discriminate temperature in time and space. The generalization potential of AVGL technique is highlighted when the AVGL–Temperature relation is computed at one region, where the temperature is known, and then used to compute the temperature change estimates in the other sub-regions. Those results are validated by finite elements simulations. According to [26], the AVGL technique is independent from the transducer, because it is based on relative intensity variations. The authors state if it is possible to obtain images that have quality enough to be used to develop a diagnosis hypothesis good quality images with a given transducer, it will be possible to apply the AVGL technique. Hence, different from techniques that demand raw ultrasound signals, the AVGL technique can be used to estimate temperature variations from conventional B-mode images.

According to the International Vocabulary of Metrology (VIM) [27], calibration “establishes a relation between the quantity values with measurement uncertainties provided by measurement standards”. Hence, independently of the ultrasonic technique employed in non-invasive evaluation of temperature variation, the calibration is necessary to ensure a reliable estimation. As previous mentioned, literature points out the MRI as the reference method for monitoring ultrasound heating treatments, achieving a temperature resolution of 0.5 °C · cm⁻³ [28]. Nevertheless, as defined in the Guide to the expression of uncertainty of measurement (GUM) [29], the resolution is just one of the uncertainty components. Hence, it is expected that, for any method, disregarding any other source of uncertainty, the expanded uncertainty would be at least the method resolution. In practice, the expanded uncertainty is in general higher than the resolution. As far as we know, the uncertainty of the MRI method was never estimated.

The models and measurements of echo-shift and CBE methods suggest its dependence on the properties of individual scatterers or scattering regions in the clinical hyperthermia temperature range [22], [26]. Hence, the calibration of the echo-shift and CBE from different tissues is indispensable to applications involving temperature estimation over different tissue types. Echo-shift techniques were calibrated to estimate the temperature in homogeneous tissue phantoms [10], [30] and in small regions where shifts can be separated from the behaviour of surrounding tissue [31], [32]. Simon et al (1998) [10] indicate that temporal echo-shifts can achieve accuracy better than 0.5 °C. Bazan et al. (2009) [11] have published resolutions in temperature of the order of tenths of °C, and a near-to-linear dependence for three ultrasonic estimation techniques (frequency, time and phase domains) applied to a homogeneous phantom with few scatterers. However, based on the GUM, resolution is just one of the sources of the measurement uncertainty estimation. Fuhrmann et al. (2016) [30] have indeed presented calibration results for the temporal echo-shifts technique according to GUM, using a phantom as study sample. The authors mention a minimum uncertainty of 12.4%, in the range from 20 °C to 44 °C, which corresponds to uncertainties between 2.5 °C and 5.5 °C. Alvarenga et al. (2015) [33] have calibrated the AVGL values extracted from B-mode images using a plastic phantom heated in a water bath from 37 °C to 46 °C. A linear relation between the AVGL and the temperature variation was observed (2.03 °C/AVGL), and an expanded uncertainty of 0.36 °C/AVGL was estimated. The AVGL values were also calibrated for a standardized tissue-mimicking material heated using physiotherapeutic ultrasound intensity levels, and it was possible to estimate the temperature from the proposed estimation models with the expanded uncertainty of 2.5 °C [34]. As far as we know, the uncertainty of AVGL technique was never estimated for in vitro tissue samples.

In this paper, the AVGL calculated from B-Mode images is assessed for non-invasive temperature estimation on in vitro porcine sample containing two different tissue types, fat and muscle. The sample is heated from 35 °C to 41 °C using a water bath, while B-mode images are acquired. A linear model is proposed relating AVGL and temperature, and the coefficients of the linear model are estimated for each tissue region studied. Moreover, the expanded uncertainty in predicting temperature using the proposed model is estimated based on the Guide to the expression of uncertainty of measurement (GUM) [29]. The innovation in comparison with the previous works [35] is that the study considers different tissue types, and the measurement uncertainty is estimated to assess the method feasibility.