A Methodology based on Formal Methods for Thermal Ablation Area

Detection

Luca Brunese, Francesco Mercaldo, Antonella Santone and Giuseppe Peter Vanoli

Department of Medicine and Health Sciences “Vincenzo Tiberio”, University of Molise, Campobasso, Italy

Keywords:

Formal Methods, Thermal Ablation, Model Checking.

Abstract:

Thermal ablation is the process related to the destruction of tissue by elevated tissue temperatures or depressed

tissue temperatures. The machine exploited to perform for this process is named thermal ablator, requiring

in input the area of the tissue to be subjected to treatment. In this proposal, with the aim to assist doctors

in the process of the detection of the area targeted by the thermal ablator, we propose a methodology based

on formal methods considering the representation of medical images in terms of formal and mathematical

representations for the detection of the area.

1 INTRODUCTION AND

BACKGROUND

The thermal ablation is deﬁned as a needle-based

treatment ﬁnalized to destroy cancerous or not nor-

mal tissue. There are two main ablation methods i.e.,

the extreme cold (also known as cryoablation) and the

extreme heat (also known as radiofrequency or mi-

crowave ablation) (Choi and Jung, 2020).

The focus of thermal ablation is the cancer tis-

sue destruction considering the generation of cyto-

toxic temperatures for a really short time-window in

a not invasive way, clearly without damaging vital

structures adjacent to the cancerous area. Typically

considered techniques to perform the thermal ablation

procedure for destroying tissue by elevating the tissue

temperature above 55

◦

C in the cancerous area include

radiofrequency, microwave, ultrasound, and also laser

ablation (Heged

us et al., 2020). The cryoablation,

from the other side, considers subzero temperatures

to selectively freeze with the aim to destroy (only) the

cancerous tissue. The innovation represents by both

these ablative procedures is that they provide a mini-

mal (e.g. percutaneously or laparoscopically) or non-

invasive approach to the tumour therapy (Zhang et al.,

2020).

The damage of the tissue can be controlled in an

accurate way by considering a range of focused ultra-

sound transducers with different sonication sizes. In

this context, medical images (for instance, magnetic

resonance and computed tomography) allows the ex-

perts to continuously monitor the temperature rise in

real time, allowing also in real-time the quantiﬁcation

of the dose of the therapy.

From the other side, ultrasound imaging and tech-

nique for the characterization of the tissue (for in-

stance, elastography) can be exploited for monitoring

the treatment relating to several clinical applications.

Depending on the equipment and parameters consid-

ered, the volume of focused ultrasound lesions can be

as small as a grain of rice (i.e., 10 cubic millimeters)

(Song et al., 2013). This allows for an extremely lo-

calized treatment and a sharp border between treated

and untreated areas.

For treatment of larger structures, as for instance

tumors, multiple lesions can be combined in order to

contain the full volume (Uchida et al., 2012; Song

et al., 2013) of the cancerous area. A cooling pe-

riod between different sonications is typically con-

sidered with the aim to reduce hte possibility of un-

wanted heating of surrounding tissue. This is the rea-

son way, the treatment of really large tissue struc-

tures can be time-consuming. Anyway, optimized

scanning algorithms, the injection of microbubbles

aimed to increase the absorption of acoustic energy,

and the adoption of spiral sonications are techniques

currently exploited ofr the reduction of the treatments

time (Brunese et al., 2019b).

The treatments exploiting thermal ablation are ob-

taining an increasing attention, as a matter of fact

are considered an alternative to classic invasive surgi-

cal therapies, with particular regards to patients with

contraindications or those who refuse open surgery

(De Baere and Deschamps, 2014).

Brunese, L., Mercaldo, F., Santone, A. and Vanoli, G.

A Methodology based on Formal Methods for Thermal Ablation Area Detection.

DOI: 10.5220/0010394801910194

In Proceedings of the 14th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2021) - Volume 3: BIOINFORMATICS, pages 191-194

ISBN: 978-989-758-490-9

191

Today, thermal ablation is exploited in clinical

applications with particular regard for kidney treat-

ing, prostate and others non-operable liver tumors.

In this direction there is also an increasing applica-

tion of thermal ablation techniques related to other or-

gan sites including the brain (Brunese et al., 2020c),

prostate, breast, lung, pancreas, thyroid and bone,

symptomatic uterine ﬁbroids; tumors in the prostate

(Blana et al., 2004; Brunese et al., 2020f), breast, low

back pain and brain disorders such as essential tremor,

disease of Parkinson and neuropathic pain. The po-

tential beneﬁts of thermal ablation therapy are includ-

ing reduced morbidity but also mortality in compari-

son with standard surgical resection and the ability to

treat patients who are not surgical candidates (Thanos

et al., 2004).

Even researchers have demonstrated that thermal

ablation represents a successful technique for reduce

tumours surface with minimal thermal damage to sur-

rounding healthy tissue(Sajjadi et al., 2011), this tech-

nique requires expert pathologist and radiologist to

localise the cancer area target of the thermal abla-

tion (Baisi et al., 2013; Morgan et al., 2010; Raveglia

et al., ).

Starting from these considerations, in this paper

we introduce a proposal for a methodology based on

formal methods for the automatic detection of the

cancer area subjected to thermal ablation.

2 MODEL CHECKING FOR

THERMAL ABLATION AREA

DETECTION

In Figure 1 we show the ﬂowchart related to our pro-

posal.

We start from the analysis of medical images,

that we convert into numerical values by exploiting

radiomics i.e., a method that extracts a large num-

ber of features from radiographic medical images us-

ing data-characterisation algorithms (Brunese et al.,

2020d; Brunese et al., 2020b; Iadarola et al., 2020;

Brunese et al., 2020a).

We consider radiomics since it has been shown

that it is able to exhibit disease characteristics that the

naked eye fails (Brunese et al., 2020a).

In detail, we exploit different radiomic features

belonging to ﬁve different categories (Van Griethuy-

sen et al., 2017):

• First Order: this category describes the distribu-

tion of voxel intensities within the ROI (i.e., the

region of interest, in this study related to the areas

in the MRI interested by the cancer);

• Shape: this feature category includes descrip-

tors of the three-dimensional size and shape of

the ROI. These features are independent from the

gray level intensity distribution in the ROI and are

therefore only calculated on the non-derived im-

age and mask;

• Gray Level Co-occurrence Matrix (GLCM): this

category considers the spatial relationship of pix-

els is the gray-level co-occurrence matrix i.e., the

gray-level spatial dependence matrix. The GLCM

functions characterize the texture of an image by

computing how often pairs of pixel with speciﬁc

values and in a speciﬁed spatial relationship occur

in an image and then extracting statistical mea-

sures from this matrix;

• Gray Level Run Length Matrix (GLRLM): the

grey-level run length matrix (GLRLM) gives the

size of homogeneous runs for each grey level. It

quantiﬁes gray level runs, which are deﬁned as the

length in number of pixels, of consecutive pixels

that have the same gray level value;

• Gray Level Size Zone Matrix (GLSZM): the fea-

tures belonging to this category quantify gray

level zones in an image. A gray level zone is de-

ﬁned as the number of connected voxels that share

the same gray level intensity. A voxel is consid-

ered connected if the distance is 1 according to the

inﬁnity norm.

The radiomic feature set obtained from each med-

ical image related to the patient under analysis, is then

translated into a formal model by exploiting the Cal-

culus of Communicating Systems (Milner, 1989). To

detect the area of the thermal ablation on the formal

model (Casolare et al., 2019; Casolare et al., 2020),

we need a set of properties expressed in a tempo-

ral logic, for instance, in the mu-calculus logic (Stir-

ling, 1989), describing the cancerous area subjected

to thermal ablation. The properties are formulated

with the knowledge of expert radiologists and pathol-

ogists. In details, our proposal considers several prop-

erties, each one related to a particular area of the med-

ical image. For example, if we ideally divide the

medical image into four equal parts, we will deﬁne

four properties, each one relating to one of the four

areas. Once obtained the formal model and the re-

lated properties, we invoke a formal veriﬁcation envi-

ronment (for instance, the CWB-NC

) to verify if the

thermal ablation area properties are satisﬁed by the

formal model obtained from the medical images.

When the formal veriﬁcation environment outputs

TRUE on a certain property, the formal model will

https://www3.cs.stonybrook.edu/

∼

cwb/

BIOINFORMATICS 2021 - 12th International Conference on Bioinformatics Models, Methods and Algorithms

192

Figure 1: The ﬂowchart.

exhibit the area subjected to thermal ablation in the

zone identiﬁed by the formula. Otherwise, the formal

veriﬁcation environment outputs FALSE meaning that

the formal model will not exhibit an area subjected to

thermal ablation in the zone identiﬁed by the formula

(Iadarola et al., 2019; Brunese et al., 2020e; Brunese

et al., 2019a).

3 CONCLUSION AND FUTURE

WORK

The thermal ablation technique is typically used for

patients with unresectable and borderline resectable

disease, which may be due to the size, number or lo-

cation of the tumors, or for patients judged inoperable

due to the poor helath of the patient. This technique

requires the aid of radiologists and pathologists to ex-

actly localise the tumour area object of the thermal ab-

lation. In this paper, we propose a method to localise

the cancerous area subjected to the thermal ablation

therapy by exploiting medical image analysis. In de-

tails, we consider radiomics features to obtain numer-

ical values from medical images, and model checking,

to automatically detect the area for the thermal abla-

tion application. As future works, we plan to specify

the temporal logic properties for the automatic area

detection. Moreover, it will be of interest to under-

stand whether the automatic area detection properties

rightly work on different type on organ. Moreover,

we plan to apply also deep techniques in order to un-

derstand the results. We plan to apply an architecture

based on Convolution Neural Network, typically ex-

ploited for processing visual data and 2D data. Basi-

cally, a CNN is made up of one or more convolutional

layers with fully connected upward layers. It also

consider common weights and layers (i.e., pooling

layers). In particular, ”max-pooling” is often used in

Fukushima’s convolutional architecture (Deng et al.,

2014), allowing CNNs to take advantage of 2D input

structures. As a matter of fact, they are particularly ef-

fective in the area of images and speech recognition.

REFERENCES

Baisi, A., De Simone, M., Raveglia, F., and Ciofﬁ, U.

(2013). Thermal ablation in the treatment of lung can-

cer: present and future. European Journal of Cardio-

Thoracic Surgery, 43(4):683–686.

Blana, A., Walter, B., Rogenhofer, S., and Wieland, W. F.

(2004). High-intensity focused ultrasound for the

treatment of localized prostate cancer: 5-year expe-

A Methodology based on Formal Methods for Thermal Ablation Area Detection

193

rience. Urology, 63(2):297–300.

Brunese, L., Martinelli, F., Mercaldo, F., and Santone, A.

(2020a). Deep learning for heart disease detection

through cardiac sounds. Procedia Computer Science,

176:2202–2211.

Brunese, L., Martinelli, F., Mercaldo, F., and Santone, A.

(2020b). Machine learning for coronavirus covid-19

detection from chest x-rays. Procedia Computer Sci-

ence, 176:2212–2221.

Brunese, L., Mercaldo, F., Reginelli, A., and Santone, A.

(2019a). Formal methods for prostate cancer glea-

son score and treatment prediction using radiomic

biomarkers. Magnetic resonance imaging.

Brunese, L., Mercaldo, F., Reginelli, A., and Santone,

A. (2019b). Radiomic features for medical images

tamper detection by equivalence checking. Procedia

Computer Science, 159:1795–1802.

Brunese, L., Mercaldo, F., Reginelli, A., and Santone,

A. (2020c). An ensemble learning approach for

brain cancer detection exploiting radiomic features.

Computer methods and programs in biomedicine,

185:105134.

Brunese, L., Mercaldo, F., Reginelli, A., and Santone,

A. (2020d). Explainable deep learning for pul-

monary disease and coronavirus covid-19 detection

from x-rays. Computer Methods and Programs in

Biomedicine, 196:105608.

Brunese, L., Mercaldo, F., Reginelli, A., and Santone, A.

(2020e). Formal methods for prostate cancer glea-

son score and treatment prediction using radiomic

biomarkers. Magnetic resonance imaging, 66:165–

175.

Brunese, L., Mercaldo, F., Reginelli, A., and Santone,

A. (2020f). Radiomics for gleason score detection

through deep learning. Sensors, 20(18):5411.

Casolare, R., Martinelli, F., Mercaldo, F., and Santone, A.

(2019). A model checking based proposal for mo-

bile colluding attack detection. In 2019 IEEE Inter-

national Conference on Big Data (Big Data), pages

5998–6000. IEEE.

Casolare, R., Martinelli, F., Mercaldo, F., and Santone, A.

(2020). Malicious collusion detection in mobile en-

vironment by means of model checking. In 2020

International Joint Conference on Neural Networks

(IJCNN), pages 1–6. IEEE.

Choi, Y. and Jung, S.-L. (2020). Efﬁcacy and safety of

thermal ablation techniques for the treatment of pri-

mary papillary thyroid microcarcinoma: a systematic

review and meta-analysis. Thyroid, 30(5):720–731.

De Baere, T. and Deschamps, F. (2014). New tumor abla-

tion techniques for cancer treatment (microwave, elec-

troporation). Diagnostic and interventional imaging,

95(7-8):677–682.

Deng, L., Yu, D., et al. (2014). Deep learning: methods

and applications. Foundations and Trends

 in Signal

Processing, 7(3–4):197–387.

Heged

us, L., Miyauchi, A., and Tuttle, R. M. (2020). Non-

surgical thermal ablation of thyroid nodules: Not if,

but why, when, and how? Thyroid.

Iadarola, G., Martinelli, F., Mercaldo, F., and Santone, A.

(2019). Formal methods for android banking malware

analysis and detection. In 2019 Sixth International

Conference on Internet of Things: Systems, Manage-

ment and Security (IOTSMS), pages 331–336. IEEE.

Iadarola, G., Martinelli, F., Mercaldo, F., and Santone, A.

(2020). Image-based malware family detection: An

assessment between feature extraction and classiﬁca-

tion techniques. In IoTBDS, pages 499–506.

Milner, R. (1989). Communication and concurrency. PHI

Series in computer science. Prentice Hall.

Morgan, G. J., Clarke, K., Caldarone, C., and Benson,

L. N. (2010). Radiolucent retractor for angiographic

analysis during hybrid congenital cardiac procedures.

The Journal of thoracic and cardiovascular surgery,

140(5):1195–1196.

Raveglia, F., Rizzi, A., De Simone, M., Ciofﬁ, U., Sacrini,

A., and Baisi, A. State of the art in alternative treat-

ments for lung cancer: Thermal ablation therapy.

Sajjadi, A. Y., Mitra, K., and Grace, M. (2011). Ablation

of subsurface tumors using an ultra-short pulse laser.

Optics and Lasers in Engineering, 49(3):451–456.

Song, J. H., Yoo, Y., Song, T.-K., and Chang, J. H. (2013).

Real-time monitoring of hifu treatment using pulse in-

version. Physics in Medicine & Biology, 58(15):5333.

Stirling, C. (1989). An introduction to modal and temporal

logics for ccs. In Yonezawa, A. and Ito, T., editors,

Concurrency: Theory, Language, And Architecture,

volume 491 of LNCS, pages 2–20. Springer.

Thanos, L., Mylona, S., Pomoni, M., Kalioras, V., Zoganas,

L., and Batakis, N. (2004). Primary lung cancer: treat-

ment with radio-frequency thermal ablation. Euro-

pean radiology, 14(5):897–901.

Uchida, T., Nakano, M., Hongo, S., Shoji, S., Nagata, Y.,

Satoh, T., Baba, S., Usui, Y., and Terachi, T. (2012).

High-intensity focused ultrasound therapy for prostate

cancer. International Journal of Urology, 19(3):187–

201.

Van Griethuysen, J. J., Fedorov, A., Parmar, C., Hosny, A.,

Aucoin, N., Narayan, V., Beets-Tan, R. G., Fillion-

Robin, J.-C., Pieper, S., and Aerts, H. J. (2017).

Computational radiomics system to decode the radio-

graphic phenotype. Cancer research, 77(21):e104–

e107.

Zhang, T., Liang, W., Song, Y., Wang, Z., and Zhang, D.

(2020). Us-ct fusion image-guided microwave abla-

tion of lung cancer—-a new mode of image guidance

in lung cancer ablation. ADVANCED ULTRASOUND

IN DIAGNOSIS AND THERAPY, 4(4):343–348.

BIOINFORMATICS 2021 - 12th International Conference on Bioinformatics Models, Methods and Algorithms

194