Registration and evaluation of thermal ablation of tumors

Registration and evaluation of thermal

ablation of tumors

Michelle Dekker (10739297)

University of Amsterdam

Academic Medical Center

Antoni van Leeuwenhoek

Netherlands Cancer Institute

Plesmanlaan 121 Amsterdam

January 2017 - May 2017

Lotte Lutkenhaus, Postdoc

Maarten Buiter, Head of Clinical Informatics Radiotherapy

Robin Langerak, Assistant Professor

ABSTRACT

Objective

The purpose of this study was to evaluate whether the quality of registrations with Mirada RTx© is sufficient to use registrations in clinical setting and serve as a visual guide for interventional radiologists during microwave ablations. Also a workflow was created.

Methods

CT-scans of patients with kidney or liver tumors treated with microwave ablation were used for the registrations. Tumors and ablation zones were contoured on the CT-scans to evaluate the registrations quantitatively by determining the Dice Similarity Coefficient and Mean Surface Distance. Qualitative evaluation was performed with a three-point scale by a radiologic technologist. The workflow was written down in an instruction document and optimized using an iterative development method.

Results

CT-scans of 22 patients (23 tumors) were analyzed. With 64 CT-scans, 59 registrations (planning – needle (PN), planning – control (PC), needle – control (NC)) could be performed. Of all

registrations, 40.7 % (24/59) resulted in a DSC of 0.70 or higher. 33.9 % (20/59) of the registrations resulted in a MSD between registered contours of less than a millimeter. Most MSDs were negative (46/59), indicating that in most cases the floating structure was bigger than the target structure. Qualitative evaluation showed that 93.2% (55/59) of the registrations could be used clinically.

Conclusions

PN registrations can be used clinically according to quantitative evaluations as well as qualitative evaluations. PN registrations could support the interventional radiologist by assessing the needle placement. PC and NC registrations can be used clinically according to qualitative evaluations, but not according to quantitative evaluations. PC and NC registrations could support the interventional radiologist by assessing the success of the treatment.

Introduction

Improvements in health care during the last centuries have led to a higher life expectancy. As a consequence, degenerative and man-made diseases have become the primary causes of morbidity and mortality (1). Uncontained growth of cells, commonly known as cancer, is such a disease.

In the Netherlands, approximately 100.000 patients are diagnosed with cancer every year, of which 90% is 50 years or older (2). Approximately 12.200 of all new yearly cases is lung cancer (3), 2.700 patients get renal cancer (2) and about 500 patients per year are diagnosed with hepatic cancer (4). For these types of cancer, a large group of patients is not eligible for surgical treatment due to several causes. For patients with renal cancer, these causes are advanced age, multiple comorbidities (which increases with age) and already altered renal function (5). The main cause for patients with hepatic cancer is an unresectable tumor (6,7) and with lung cancer, the reason is poor cardiorespiratory reserve (6).

A good alternative for these patients who cannot be operated is local ablative treatment. Ablation consists of heating the tumor to a temperature at which tumor cells die. When the heating is caused by microwaves, it is called microwave ablation (MWA). During this procedure, an interventional radiologist places a needle in the tumor using CT-guidance. An antenna that runs through the needle is heated during the treatment. In turn, the needle heats the tissue until it reaches 80 degrees Celsius, causing the tumor cells to die. In comparison to surgery, MWA has benefits like shorter hospitalization, fewer and less severe complications, faster recovery, less pain, preservation of normal tissue, avoidance of general anesthesia and the ability to repeat the treatment (8). The results of this treatment and surgery are alike (9).

During the MWA procedure, the interventional radiologist uses different CT-scans for several reasons as shown in figure 1. The first CT-scan is acquired at the start of the treatment, and is used to identify the location of the tumor and to determine where the needle needs to be placed. It is therefore called the planning CT-scan. The second CT-scan is acquired after the needle has been placed. This needle CT-scan is used to assess whether the needle has been placed correctly. After treatment, the last CT-scan, the control CT-scan, is acquired to evaluate whether the entire tumor has been treated. This is all based on visual comparison of the CT-scans that are acquired during the procedure.

Figure 1. Current use of CT-scans during the procedure.

planning

CT-scan

• Locate the

tumor

• Determine

needle

position

Needle CT-scan

• Assesment of

needle

placement

Control CT-scan

• Evaluate if

the entire

tumor has

been treated

Comparison of CT-scans can be difficult due to the fact that the patient may move and breathe during the procedure, causing differences in the obtained images (10). It is also possible that the tumor is not visible on one of the scans. As a consequence, the interventional radiologist mentally maps the location of the tumor from one scan to the other. This procedure is inconvenient, time-consuming, and potentially inaccurate. Apart from the intuition of the interventional radiologist, there currently is no other guidance during the treatment (11) and therefore the quality of the treatment depends not only on the location and size of the tumor but also on the experience of the interventional radiologist (11– 13). Although the recurrence rate of MWA treatments is low (14,15), treatment is not optimal yet. In our institute, 9 out of 127 ablations performed in 2016 were re-ablations. As far as we could assess, these re-ablations could have been prevented if the first ablation would have been successful. Cases where parts of the tumor remain because the interventional radiologist had difficulty assessing the location of the tumor or needle properly should be prevented. To reduce the dependency on the experience of the interventional radiologist, the CT-scans can be registered (projected) onto each other. The registration may enable the interventional radiologist to see if the position of the needle needs to be adjusted. This way, registrations visually support the interventional radiologist and could potentially benefit the procedure.

Although registration is currently not used during MWA treatments, clinical software is available which can perform registrations. This software is called Mirada RTx©. The aim of this research is to evaluate the quality of registrations with Mirada RTx software and to examine in which way registrations can be used to support the comparison of CT-scans and make treatments with microwave ablation more independent of the experience of the interventional radiologist.

Background

Registration

The ultimate goal of image analysis is to extract useful information from the processed images (16). Comparing two images is much easier when the images are spatially aligned. This way, the observer does not have to map the images mentally onto each other. Also, inter-observer variability reduces because the images are mapped the same way for every observer. The spatial alignment is created through finding similarities between two images. This process is called image registration (17). Images can be acquired at different times and with other modalities. Together with the type of registration these aspects determine the sort of registration.

Type of registration

When a registration is performed, one image is projected onto the other image. The image which does not move is referred to as the target; the other image is called the floating image. There are several ways to determine how the floating image should be projected on the target. First of all, the floating image could be rotated and translated. When only those two actions are performed, the registration is called rigid (18,19). When scaling is added, the registration is called affine. A projective registration also drops the restriction of parallelism (18). The last type of registration is called curved, because lines can be mapped to curves (18,19). Curved registrations are also referred to in literature as deformable, elastic or fluid registrations (20).

Scale

The last aspect to mention is the scale of the registration. When the whole floating image is being mapped onto the target image, the registration is global. When only a part of the floating image, a region of interest (ROI), is taken into account, it is a local registration (18). Every mode of registration mentioned above can be performed on a global as well as on a local scale, as shown in figure 2.

Clinical setting

Registrations are applied in multiple areas of the medical field for various reasons. One reason to perform a registration is to combine information gathered from different modalities, for example from CT and MRI. It is necessary to register these images because different modalities vary in field of view, resolution, and slice orientation (21). Another reason to perform a registration is to correct for movement of the patient during an examination or because the patient cannot be repositioned in exactly the same way as the previous visit (21).

An example within the medical field where registrations are used is computer-aided diagnosis (20). A commonly used combination for this purpose is a registration between a scan and a PET-scan. CT-scans provide anatomic details, while PET-CT-scans are used for the functional specificity. These two modalities therefore complement each other and compose a complete overview of the information needed for diagnosis. This combination is nowadays so common that since several years these registrations are not only performed on software level, but also on hardware level. This is achieved by combining both modalities into one scanner (22).

Methods

To evaluate the quality of registrations in Mirada and to search for a workflow, CT-scans of patients treated with MWA between June and December 2016 were used. Only patients with a liver or kidney tumor were selected. The procedures were performed at the Antoni van Leeuwenhoek hospital by interventional radiologists. The scans were categorized into three types of scans: planning-CTs, needle-CTs and control-needle-CTs. In some cases, a second needle-CT is acquired after ablation, to perform additional ablation in an attempt to make sure the ablation is radical. In this case, the second needle-CT shows a partial ablation zone as opposed to the first needle CT which only shows tumor.

Registrations

First, tumors were contoured on planning-CTs and needle-CTs. Ablation zones (i.e. the volume of ablated tissue) were contoured on control-CTs. Delineations were performed by a single observer and approved by an interventional radiologist. These steps are referred to as phase 1 in figure 3.

Next, registrations of 1) needle-CTs on planning-CTs, 2) control-CTs on planning-CTs and 3) control-CTs on needle-CTs were performed. Registrations were executed rigid to avoid deformation of the needle and a distorted representation of the ratio between the tumor and ablation zone. Two local registrations were performed per set of scans with a three-dimensional box around the tumor and a part of the kidney or liver. In Mirada, rigid registrations can be performed on medium, fast, or fine level; all registrations were performed on the fine level. These steps are referred to as phase 2 in figure 3. Finally, the registrations were quantitatively and qualitatively evaluated. For the quantitative analysis, the Dice Similarity Coefficient (DSC) and the Mean Surface Distance (MSD) were calculated for each registration.

With the volumes of the contours and the intersections, the Dice Similarity Coefficient (DSC) was calculated. The DSC is a representation of the similarity between two structures, where 0 stands for no similarity and 1 for perfect similarity. According to literature, a DSC of 0.70 or higher indicates excellent agreement (23) and is therefore considered to be clinically acceptable. The DSC is calculated using the following equation,

𝐷𝑆𝐶(𝑇, 𝐹) = 2_{|𝑇|+|𝐹|}|𝑇∩𝐹| (1),

where T is the volume of the target structure and F is the volume of the floating structure, both in cm3. The Mean Surface Distance (MSD) represents the mean distance between corresponding points on two contours. In this study, the MSD of 100 points was calculated using the following equation,

𝑀𝑆𝐷(𝑇, 𝐹) = _(𝑛 1

𝑇+ 𝑛𝐹)(∑ 𝑑𝑖

𝑛𝑇

𝑖=1 + ∑𝑛𝑗=1𝐹 𝑑𝑗) (2)

, where T stands for the target structure and F for the floating structure. NT and nF represent the number

of points on the two corresponding contours and di, dj are the closest distances from each point on the

target contour to the floating contour.

Also, registrations were analyzed qualitatively with a three-point scale by a radiologic technologist, who performs registrations on a daily basis. The registrations could be judged as bad, moderate or good. Bad meaning not good enough to be clinically used, moderate meaning can be used clinically but could be better, and good can be used clinically. Registrations which were scored as ‘bad’, were discussed with the radiologic technologist and potentially re-registered.

Figure 3. Phases of the registration process.

Workflow

In order to assess the current workflow of the MWA procedure, MWA-treatments were attended every Monday. During these moments, the focus of the observation was to find out at which moments the interventional radiologist needs support and what kind of tools are used to evaluate the scans and perform the treatment.

Every two weeks a meeting was organized where ideas for the use of the program could be discussed. The results of these observations and meetings were processed into a customized layout of Mirada and a workflow. The workflow was written down in an instruction document and optimized using an iterative development method. Instructions were tested two times using the ‘think aloud’ method. The first evaluation was to test if the instructions were clear, this evaluation was performed with someone who never used Mirada before. Feedback from this evaluation was processed and a second version of the instruction document was made. This version was tested with two radiologic technologists who assist the interventional radiologists during the MWA treatments. The results from this testing were used to optimize the layout and workflow to a final version.

Materials

For the performance of the registrations, Mirada Medical RTx Advanced 1.6© was used. To compute the DSC and MSD in-house made software was used. Lastly, statistical analysis was performed with IBM SPSS Statistics version 22©.

Phase 1

•Contour the tumor/ablationzone •Check interventional radiologist

Phase 2

•Register the scans in Mirada •Compute the intersection

Phase 3

•Quantitative analysis •Qualitative evaluation

Results

CT-scans of 22 patients with a total of 23 tumors were analyzed. Of the 64 CT-scans, 23 were planning-CTs, 19 were needle-planning-CTs, and 22 were control-CTs. Of all 59 registrations, 19 consisted of a needle-CT registered to a planning-CT (PN), 22 registrations were between a control-CT and a planning-CT (PC), and the remaining 18 registrations were between a control-CT and needle-CT (NC). Table 1 shows characteristics of the treated patients and the tumors, figure 4 shows the different types of registration for a typical patient.

Patients (n) 22

Age (years) (mean ± SD) 69.6 ± 7.0

Sex (n) Male 16 Female 6 Tumors (n) Kidney 17 Liver 6

Ablation time (min) (mean ± SD) 8.4 ± 5.1

Size of tumor (cm3) (mean ± SD) 23.7 ± 36.4

Table 1. Baseline Characteristics of the treated patients and the tumors.

Figure 4. Three registrations of CT-scans obtained during a MWA treatment of a kidney tumor. Left, registration of

planning-CT (red contour1) and needle-CT (blue contour1). In the middle, registration of planning-CT (red contour1) and control-CT (green contour1). Right, registration of needle-CT (blue contour1) and control-CT (green contour1).

1

Registrations

Of all registrations, 40.7 % (24/59) resulted in a DSC of 0.70 or higher. In figure 4 the DSC values are shown per registration type. The PN registrations have a median DSC of 0.78 (range 0.49 - 0.86) (figure 5a). When needle-CT scans acquired after ablation (n = 10) are excluded from this analysis, the range in DSC reduces to 0.69 – 0.86 (figure 5b). The range in DSC for the PC (0.00 – 0.86) and NC (0.35 – 0.87) registrations stays the same without needle-CTs acquired after ablation. However, omitting the registrations with needle-CTs acquired after ablation reduces the difference between the median DSC of PC and NC from 0.22 to 0.05.

Figure 5a (left) and 5b (right). Dice Similarity Coefficient values of the registrations per registration type: planning

-needle (PN), planning - control (PC) and -needle - control (NC). a) results from all 59 registrations. b) results after excluding needle-CTs acquired after ablation (n = 49).

Of all registrations, 33.9 % (20/59) resulted in a MSD between registered contours of less than a millimeter. Most MSDs were negative (46/59), indicating that in most cases the floating structure was bigger than the target structure. The PN registrations have a median MSD of -0.7 mm (range -7.1 – 0.3 mm) (figure 6a). When needle-CT scans acquired after ablation (n = 10) were excluded from this analysis, the range in MSD between registered contours reduced to -3.2 – 0.3 mm (figure 6b). Omitting the NC registrations where the needle-CT contains ablation zone increased the range in MSD of NC registrations from -4.2 – 1.7 mm to -4.7 – 4.1 mm.

Figure 6a (left) and 6b (right). Mean Surface Distance (in mm) of the registrations per registration type: planning

-needle (PN), planning - control (PC) and -needle - control (NC). a) results from all 59 registrations. b) results after excluding needle-CTs acquired after ablation (n = 49).

Qualitative evaluation of the registrations showed that 93.2% (55/59) of the registrations could be used clinically (table 2). Four registrations were judged as ‘bad’ and were further discussed. Two of these four were re-registered with a larger box, after wich the registrations were re-evaluated as ‘good’. The other two registrations were from the same patient, who had hadmajor bleeding during the procedure. This complication might explain why the images do not match, and could therefore not be registered correctly.

Bad Moderate Good Total

Planning - Needle 2 3 14 19 Planning - Control 1 5 16 22 Needle - Control 1 1 16 18 Total 4 9 46 59

Table 2. Qualitative evaluation of the registrations.

Workflow

The observations of in total 49 MWA procedures (4 lung, 6 liver and 39 kidney), spread out over 18 treatment days, showed that the interventional radiologists do not use many tools to evaluate the CT-scans. They use the CT workplace to visually assess the scans and sometimes another computer to view a diagnostic scan made weeks or months before the procedure. Further, sometimes a tool is used to draw a line which represents how the interventional radiologist wants to position the needle and to measure if the needle is long enough to reach the desired deepest point.

The first evaluation was to test if the instructions were clear by letting someone who never used Mirada before (n=1), perform the steps from the instruction document. The overall impression of this

person was that the instructions were clear, although some steps needed some further explaining. The biggest confusion arose while placing the registration box. Based on this evaluation four adjustments were made in the first version of the instruction document. The total time needed to go through all the steps was eight and a half minute.

The second evaluation was with the radiologic technologists (n=2). Both were enthusiastic about the program and saw what the added value of the use of registrations during treatments could be. It took the first radiologic technologist eleven and a half minute to complete all the steps, the other radiologic technologist sixteen and a half minute. Both had the most trouble with finding the buttons and doubted if they performed the correct actions because they had little or no explanation about the aim of the steps. Based on this evaluation four adjustments were made in the second version of the document. The final version of the instruction document can be found in appendix 1.

Discussion

In this study, we showed that the quality of local rigid registrations is sufficient to use in a clinical setting to serve as visual guidance for an interventional radiologist during MWA. In addition, we proposed a workflow for the use of Mirada during MWA treatments.

In our study, we found an average DSC of 0.60 and an average MSD of -1.80 mm. Luu et al. evaluated registrations of diagnostic CT images to intra-operative CT images for CT guided liver radiofrequency ablation (RFA) with the DSC and MSD (11). They found a DSC of 0.91 and MSD of 5.3 mm of the liver surfaces. Compared to our study, their DSC value is higher and therefore better, while their MSD is greater and therefore worse. The difference in DSC and MSD can be explained because they 1) used non-rigid image registration and 2) registered the diagnostic scan with the needle-CT. Because the registrations were non-rigid, the overlap between the two structures was bigger, resulting in higher DSC values. Because the scans were acquired at different moments, the position of the patient was not exactly the same and the liver had a different shape. This results in a greater distance between the position of the points on the various scans.

In another study, Luu et al. performed rigid registrations with pre- and post-interventional CT images for assessing treatment success in liver RFA treatment (24). They found an average DSC of 0.879 and a MSD of 5.53 mm of the liver surfaces. Again, our DSC value is lower and therefore worse, while our MSD is lower and therefore better. The difference in MSD can be explained with the same reason as for the difference between Luu et al. (11) and our study. Difference in obtained DSC values is explained by the fact that Luu et al. looked at the overlap between surfaces of the whole liver, while our study focused on the overlap between the tumor and ablation zone. Relatively speaking the change of volume between the two scans is much bigger when only tumor and ablation zone are considered. Despite the differences in results, both studies conclude, in accordance with this study, that registration is a promising tool to improve the treatment.

Lee et al. looked at the DSC, MSD and Hausdorff Distances (HD) of registrations between a set of reference (atlas) images per organ in the abdomen (25). One of the used registration tools performed rigid registrations. When the DSC of these registrations are compared to our study, the DSC of the kidney registrations are higher in our study, whereas the DSC of the liver registrations are lower. But, thenumber of liver registrations in our study was very low (n=12).

We evaluated the quality of the registrations quantitatively and qualitatively, since no clinical standard exists for the evaluation of registrations (26). The region of interest was the tumor/ablation zone, which is not a fixed number of voxels in the image and therefore needed to be manually delineated. For quantitative assessment, different kinds of similarity measures were calculated, which were based on the manually created contours. These delineations are subject to inter- and intraobserver variability, but to increase the accuracy of the contours, they were checked by an interventional radiologist as suggested by Roy et al. (27). Intra-observer variability is reportedly small (28,29), therefore we expect the resulting error to be minimal. Inter-observer variability was minimized because all the contours were drawn by the same person and checked by the same interventional radiologist. Lastly, the accuracy of the contours could also be affected by disruptions of the images caused by artifacts. These can be produced by the needle (metal artefact), which in turn causes beam hardening, scatter effects and Poisson noise. With increased noise, low contrast soft tissue boundaries, as between tumor and healthy tissue, may be obscured (30). However, if the needle is present in both scans it could be an advantage because the needle is easy to recognize and register.

Uncertainty about the correctness of the contours is not the only factor that affects the outcomes of the quantitative evaluation. In some cases, we observed that the tumor shrinks during the treatment. The shrinking effect of ablation treatments has been reported before (31), but was never quantified. Further research on this topic is therefore required. Also, a part of the kidney or liver is being ablated during the treatment to create a safety margin. According to this, we can expect to find a contour on the control-CT that does not match exactly with that on the planning-CT. Therefore, we conclude that the DSC and MSD are not useful evaluation tools for registrations of control-CTs on planning-CTs or control-CTs on needle-CTs.

Using registrations during MWA treatments can serve as a visual guide for the interventional radiologist. It can provide the interventional radiologist with a tool to improve needle placement accuracy. Also, it could be useful in estimating the location and size of the ablation zone. Lastly, it could improve the assessment of the safety margin (10). These are just the applications with CT-scans obtained during the treatment itself, registrations with a diagnostic scan or follow-up scan like in the studies of Luu et al. could also be useful. On a diagnostic scan tumors are usually better visible and could therefore make it easier to locate the tumor. Registrations with follow-up scans could show to what extent the ablation zone changes after the treatment. In addition, registrations also provide a steady basis for taking and defending decisions for the interventional radiologist.

Conclusion

Registrations

Registrations of needle-CTs on planning-CTs (PN) can be used clinically according to both quantitative and qualitative evaluations. This means that PN registrations can be used to support the interventional radiologist by assessing the needle placement.

The quality of the registrations between planning-CTs and control-CTs (PC) cannot be evaluated quantitatively, due to the inherently different volumes that are used for analysis. However, qualitative analysis showed that the scans can be registered adequately, which could provide feedback during ablation about the success of the procedure in the future. The same conclusion applies for registrations between needle-CTs and control-CTs (NC).

Workflow

A workflow to use Mirada during MWA treatments has been proposed. Both radiologic technologists could perform all the steps, and could estimate how the registrations could contribute to the treatment. However, how practical this method is, will have to be proved when the program is used in clinical setting.

In conclusion, software for registering CT-scans is available. The quality of these registrations is sufficient to use them in clinical setting, and a workflow to perform the registrations has been proposed. Now it is time to move on from words to actions and prevent the cases where parts of the tumor remain because

References

1. Abdel RO. The Epidemiologic Transition: A Theory of the Epidemiology of Population Change. Milbank Q [Internet]. 2005 [cited 2017 Mar 8];83(4):731–57. Available from:

http://www.planchet.net/EXT/ISFA/1226.nsf/769998e0a65ea348c1257052003eb94f/ea9d1a60f 8a81e9cc12576000033dddd/$FILE/Omran_2005(1971).pdf

2. KWF. KWF en nierkanker - KWF Kankerbestrijding [Internet]. [cited 2017 Mar 8]. Available from: https://www.kwf.nl/kanker/kwf-en-nierkanker/Pages/default.aspx

3. KWF en longkanker - KWF Kankerbestrijding [Internet]. [cited 2017 Mar 8]. Available from: https://www.kwf.nl/kanker/longkanker/Pages/default.aspx

4. Leverkanker [Internet]. [cited 2017 Mar 8]. Available from:

https://www.kanker.nl/bibliotheek/leverkanker/wat-is/485-leverkanker

5. Genson P-Y, Mourey E, Moulin M, Favelier S, Lucy DM, Olivier C, et al. Image-guided

percutaneous microwave ablation of small renal tumours: short- and mid-term outcomes. Quant Imaging Med Surg. 2015;5(5):649–55.

6. Simon CJ, Dupuy DE, Mayo-Smith WW. Microwave Ablation: Principles and Applications. Oncol Interv [Internet]. 2005 [cited 2017 Mar 7]; Available from:

http://pubs.rsna.org/doi/pdf/10.1148/rg.25si055501

7. Martin RCG, Scoggins CR, McMasters KM. Safety and Efficacy of Microwave Ablation of Hepatic Tumors: A Prospective Review of a 5-Year Experience. Surg Oncol. 2010;17:171–8.

8. Georgiades C, Rodriguez R. Renal Tumor Ablation. Tech Vasc Interv Radiol [Internet]. 2013 [cited 2017 Mar 9];16:230–8. Available from: http://ac.els-cdn.com/S108925161300067X/1-s2.0-

S108925161300067X-main.pdf?_tid=5dcfdf50-04a1-11e7-8f43-00000aacb361&acdnat=1489047857_48f462cba04dc077e69368f906e0c3d4

9. Katsanos K, Mailli L, Krokidis M, McGrath A, Sabharwal T, Adam A. Systematic Review and Meta-Analysis of Thermal Ablation Versus Surgical Nephrectomy for Small Renal Tumours. Cardiovasc Intervent Radiol [Internet]. 2014;37(2):427–37. Available from:

http://download.springer.com/static/pdf/649/art%253A10.1007%252Fs00270-014-0846- 9.pdf?originUrl=http%3A%2F%2Flink.springer.com%2Farticle%2F10.1007%2Fs00270-014-0846-9&token2=exp=1489048970~acl=%2Fstatic%2Fpdf%2F649%2Fart%25253A10.1007%25252Fs002 70-014-084

10. Kim KW, Lee JM, Klotz E, Kim SJ, Kim SH, Kim JY, et al. Safety margin assessment after

radiofrequency ablation of the liver using registration of preprocedure and postprocedure CT images. Am J Roentgenol. 2011;196(5).

11. Luu HM, Klink C, Niessen W, Moelker A, Van Walsum T. Non-rigid registration of liver ct images for ct-guided ablation of liver tumors. PLoS One. 2016;11(9).

12. Hildebrand P, Leibecke T, Kleemann M, Mirow L, Birth M, Bruch HP, et al. Influence of operator experience in radiofrequency ablation of malignant liver tumours on treatment outcome. Eur J Surg Oncol [Internet]. 2006 May [cited 2017 Mar 9];32(4):430–4. Available from:

http://linkinghub.elsevier.com/retrieve/pii/S0748798306000321

13. Wood BJ, Locklin JK, Viswanathan A, Kruecker J, Haemmerich D, Cebral J, et al. Technologies for Guidance of Radiofrequency Ablation in the Multimodality Interventional Suite of the Future. [cited 2017 Mar 9]; Available from:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2555973/pdf/nihms36431.pdf

14. Pavlovich CP, Walther MM, Choyke PL, Pautler SE, Chang R, Linehan WM, et al. PERCUTANEOUS RADIO FREQUENCY ABLATION OF SMALL RENAL TUMORS: INITIAL RESULTS. [cited 2017 Apr 12]; Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4136670/pdf/nihms-36471.pdf 15. Groeschl RT, Pilgrim CHC, Hanna EM, Simo KA, Swan RZ, Sindram D, et al. Microwave Ablation for

Hepatic Malignancies: A Multiinstitutional Analysis. Ann Surg [Internet]. 2013;259(6):1. Available from: http://content.wkhealth.com/linkback/openurl?sid=WKPTLP:landingpage&an=00000658-900000000-98301%5Cnpapers3://publication/doi/10.1097/SLA.0000000000000234

16. El-Gamal FE-ZA, Elmogy M, Atwan A. Current trends in medical image registration and fusion. Egypt Informatics J. 2016;17:99–124.

17. Crum WR, Hartkens T, G Hill DL. Non-rigid image registration: theory and practice. [cited 2017 Mar 9]; Available from:

http://eecs.vanderbilt.edu/courses/cs359/other_links/papers/2004_crum_harkens_hill_nonrigid _survey.pdf

18. Maintz JBA, Viergever MA. A survey of medical image registration. Med Image Anal [Internet]. 1998 [cited 2017 Mar 9];2(1):1–36. Available from:

http://s3.amazonaws.com/academia.edu.documents/7525487/10.1.1.110.3014.pdf?AWSAccess KeyId=AKIAIWOWYYGZ2Y53UL3A&Expires=1489053786&Signature=7DJND0UkkWLSKbGyw8t7N 8D4nzQ%3D&response-content-disposition=inline%3B

filename%3DA_survey_of_medical_image_registra

19. Pluim JPW, Maintz JBA, Viergever MA. Mutual-Information-Based Registration of Medical Images: A Survey. IEEE Trans Med Imaging [Internet]. 2003 [cited 2017 Mar 9];22(8). Available from: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.197.6513&rep=rep1&type=pdf 20. Oliveira FPM, Tavares JMRS. Medical image registration: a review. Comput Methods Biomech

Biomed Engin [Internet]. 2014;17(2):73–93. Available from:

http://www.ncbi.nlm.nih.gov/pubmed/22435355%5Cnhttp://www.tandfonline.com/doi/abs/10. 1080/10255842.2012.670855

21. Hill DLG, Batchelor PG, Holden M, Hawkes DJ. Medical image registration. Phys Med Biol [Internet]. 2001 [cited 2017 Mar 9];46:1–45. Available from:

http://research.mholden.org/publications/hill2001pmb.pdf

22. Townsend DW. Dual-Modality Imaging: Combining Anatomy and Function*. J Nucl Med. 2008;49(6):938–55.

23. Zijdenbos AP, Dawant BM, Margolin RA, Palmer AC. Morphometric Analysis of White Matter Lesions in MR Images: Method and Validation. IEEE Trans Med Imaging [Internet]. 1994 [cited 2017 Mar 27];13(4). Available from:

https://pdfs.semanticscholar.org/1fb5/4a3a394f48eaf757b213a78591995c499cdc.pdf 24. Luu HM, Niessen W, van Walsum T, Klink C, Moelker A. An automatic registration method for

pre- and post-interventional CT images for assessing treatment success in liver RFA treatment. Med Phys [Internet]. 2015;42(9):5559–67. Available from:

http://scitation.aip.org/content/aapm/journal/medphys/42/9/10.1118/1.4927790

25. Lee CP, Xu Z, Burke RP, Baucom RB, Benjamin K, Abramson RG, et al. Pitfalls and Opportunities with Soft Anatomy. 2015;

26. Tang M, Chen F. A Qualitative Meta Analysis Review on Medical Image Registration Evaluation. Procedia Eng [Internet]. 2012 [cited 2017 Mar 16];29:499–503. Available from:

www.sciencedirect.com

27. Roy AEF, Wells P. Volume definition in radiotherapy planning for lung cancer: How the radiologist can help. Cancer Imaging. 2006.

28. Hopper K 0, Kasales CJ, Slyke MA Van, Schwartz TA, Tenhave TR, Jozefiak JA. Analysis of Interobserver and lntraobserver Variability in CT Tumor Measurements. AJR [Internet]. 1996 [cited 2017 Mar 14];167. Available from:

http://www.ajronline.org/doi/pdf/10.2214/ajr.167.4.8819370

IMAGES. [cited 2017 Mar 14]; Available from:

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.709.8003&rep=rep1&type=pdf 30. Boas FE, Fleischmann D. CT artifacts: Causes and reduction techniques. Imaging Med [Internet].

2012 [cited 2017 Mar 14];4(2):229–40. Available from:

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.259.8500&rep=rep1&type=pdf 31. Sommer CM, Sommer SA, Mokry T, Gockner T, Gnutzmann D, Bellemann N, et al. Quantification

of Tissue Shrinkage and Dehydration Caused by Microwave Ablation_ Experimental Study in Kidneys for the Estimation of Effective Coagulation VolumeQuantification of Tissue

Shrinkage/Dehydration in MW Ablation. J Vasc Interv Radiol [Internet]. 2013 [cited 2017 Mar 21];24:1241–8. Available from:

Appendix 1.

Final version of the instruction document for the workflow in Mirada.

1

Open Mirada by clicking on 2

Find the right patient by entering the ID at

3 Double-click the patient, double-click the appropriate examination, select the first scan to open (click) and the second scan (ctrl + click)

4

Click on

5 Set review mode to ‘ Fusion and Contouring’ and drag the second scan (the one with the latest point of time) to the box ‘CT’ and finally click ‘OK’

Mark region of interest

6 _{Open the structureset MWA: a) click on b) choose MWA} 7

Select the structure ‘Registrationzone’ and click on

8 Draw on the plan CT (upper series) the area within which the entire tumor is located and a small part of the kidney / liver, check whether the tumor is in the area for all three views

ATTENTION: You can make the square bigger or smaller by dragging on one of the corners if the square in the corner is white, otherwise a new square will be drawn.

Registration 9

Click on 10

Click on and choose ‘Unlock ..’

11 Click on 'Manual Rigid' and place the purple scan, as good as possible, on the green scan by sliding or rotating (for rotating locate the mouse on the green circle)

ATTENTION: Click once again on ‘Manual Rigid’ when you are finished 12

Select the registrationzone, next click on 13

Click on the box with 3 dots and choose ‘Fine’. Wait

14

Check the registration, after successful registration: click on ; after failed registration: repeat step 11 tm 14

Icons and buttons

Ruler

Ruler in 2 directions

Manually enlarge or decrease the structure

C Lets the chrosshair appear / disappear

F11 Switch between full screen and edit

mode

Ctrl + left mouse button Move the scan