• Users Online: 2948
  • Print this page
  • Email this page

 Table of Contents  
ORIGINAL ARTICLE
Year : 2021  |  Volume : 54  |  Issue : 3  |  Page : 91-96

Laboratory virtual instrument engineering workbench-based semi-automated measurement of cranial asymmetry


1 Section of Neurosurgery, Department of Surgery, Ditmanson Medical Foundation, Chia-Yi Christian Hospital; Department of Biotechnology, Asia University, Taichung City, Taiwan
2 Division of Neurosurgery, Department of Surgery, Sijhih Cathay General Hospital; Department of Medicine, School of Medicine, Fu Jen Catholic University, New Taipei City, Taiwan
3 Division of Neurosurgery, Department of Surgery, Sijhih Cathay General Hospital; Department of Medicine, School of Medicine, Fu Jen Catholic University, New Taipei City; Department of Neurological Surgery, Tri-Service General Hospital, National Defense Medical Center, Taipei, Taiwan

Date of Submission05-May-2020
Date of Decision17-Aug-2020
Date of Acceptance11-Jan-2021
Date of Web Publication12-Jun-2021

Correspondence Address:
Cheng-Ta Hsieh
Division of Neurosurgery, Department of Surgery, Sijhih Cathay General Hospital, No. 2, Lane 59, Jiancheng Road, Xizhi District, New Taipei City 221
Taiwan
Login to access the Email id

Source of Support: None, Conflict of Interest: None


DOI: 10.4103/fjs.fjs_64_20

Rights and Permissions
  Abstract 


Background: Cranial asymmetry has been associated with the laterality of neurological diseases including chronic subdural hematoma, subdural hygroma, and stroke. Although picture archiving and communication systems (PACS) are commonly used radiologic tools, simple angle or contoured area measurement cannot accurately reflect the actual severity of cranial asymmetry. Therefore, we developed an objective semi-automated image analysis tool based on the Laboratory Virtual Instrument Engineering Workbench (LabVIEW) system and compared its efficacy and variability with those of PACS in measuring cranial asymmetry.
Methods: This image analysis software was developed on the basis of the LabVIEW system. Three sizes of plastic water pipes and computed tomographic images of the brain from three patients were used for experimental and clinical validations, respectively. We compared the percent error of the calculated areas of the pipes as well as coefficient of variation (CV) and cranial index of symmetry (CIS) ratio obtained from LabVIEW and PACS.
Results: Experimental validation showed the overall mean difference of actual size versus estimated size obtained using PACS and LabVIEW-based image analysis to be 7.51% and 4.68%, respectively. This result indicated that LabVIEW-based image analysis provided an estimated area closer to that of the actual size of the phantom, with significantly low inter- and intraobserver variability (P < 0.001). Clinical validation also showed lower variability in the CVs and CIS ratios of areas estimated using LabVIEW-based image analysis, ranging from 0.01% to 0.13% and from 89.2% to 99.1%, respectively.
Conclusion: Our study demonstrated through experimental and clinical validation that LabVIEW-based image analysis is a convenient and effective method for investigating cranial asymmetry. This imaging tool can provide more clues in understanding cranial asymmetry.

Keywords: Computed tomographic scan, cranial asymmetry, laboratory virtual instrument engineering workbench system, semi-automated analysis


How to cite this article:
Sun JM, Huang CT, Hsieh CT. Laboratory virtual instrument engineering workbench-based semi-automated measurement of cranial asymmetry. Formos J Surg 2021;54:91-6

How to cite this URL:
Sun JM, Huang CT, Hsieh CT. Laboratory virtual instrument engineering workbench-based semi-automated measurement of cranial asymmetry. Formos J Surg [serial online] 2021 [cited 2021 Sep 24];54:91-6. Available from: https://www.e-fjs.org/text.asp?2021/54/3/91/318215




  Introduction Top


Cranial asymmetry has always been an interesting topic in neurological research.[1],[2] Cranial vault asymmetry, based on the diagonal difference known as oblique diagonal difference or transcranial difference, has been commonly used to calculate the severity of cranial asymmetry among infants.[3] The most common types of cranial asymmetry are plagiocephaly, brachycephaly, or dolichocephaly. The severity of such asymmetry influences the outcome of cranial remolding treatments.[3],[4] Zonenshayn et al. introduced an objective semi-automated image analysis that calculated the cranial index of symmetry (CIS) by denoting the nasion and inion on the head.[5] The rest of the analysis was automated and calculated on the area of interest. The preliminary results showed that this objective measurement could allow the grading of the severity of positional plagiocephaly. However, no study has reported the further application of this method.

Laterality of chronic subdural hematomas (CSDH) or traumatic subdural hygromas among adults is reported to be associated with cranial asymmetry.[6],[7] Cranial asymmetry was measured using a simple method of calculating angular differences made by three lines passing through the midline and both sides of the cranium.[6] However, differences in the angles may not exactly reflect cranial asymmetry because the points are chosen manually. Currently, these intracranial parameters are measured using the picture archiving and communicating system (PACS) to investigate cranial asymmetry. However, bias may exist because the contour of the area is manually chosen. To eliminate this issue and obtain more accurate measurements, in this study, we aimed to develop an objective semi-automated tool based on the laboratory virtual instrument engineering workbench (LabVIEW) system, similar to the concept introduced by Zonenshayn et al.,[5] to evaluate cranial asymmetry using computed tomography (CT) scan images of the brain. Experimental and clinical validations are also presented.


  Materials and Methods Top


Laboratory Virtual Instrument Engineering Workbench -based image analysis

The image analysis software used in our study was developed on the basis of the LabVIEW system (National Instruments, Texas). CT images were loaded into this system, and cranial asymmetry was analyzed, as described by Zonenshayn et al.[5] The flowchart of the image analysis method is shown in [Figure 1]. The line passing through the nasion and inion, denoted manually by black spots, was considered the midline. Subsequent analysis was automated and comprised skull bone subtraction, encoding the number of pixels, and determining each cranial area. With the midline as the axis, one side of the cranium was horizontally flipped onto the other to identify overlapping and nonoverlapping areas. The CIS ratio was then calculated as twice the overlapping area divided by the total area.
Figure 1: Flowchart of image analysis

Click here to view


The first step is to determine the midline of the loaded CT image to divide the brain into left and right sides. After the user sets the “start” and “end” points on the nasion and inion, the midline that runs parallel to the long axis of the skull bone automatically connects with the points in the image analysis program [Figure 2]. A two-dimensional image subset (n × n) is also extracted from the original image where n corresponds to the length and the center corresponds to the midpoint of the midline.
Figure 2: The midline is set (red line) between the points of nasion and inion in the loaded image

Click here to view


Since cranial asymmetry is evaluated by referring to the image flipped on the midline, the loaded image is automatically rotated to let the midline to become exactly vertical. [Figure 3] shows the image before and after rotation in the second step of the image analysis.
Figure 3: Image subset before (left) and after (right) automatic rotation

Click here to view


In the third step, both the cranial and skull bone areas are segmented in the rotated image using the “nearest-neighbor-visiting” method. First, the boundary of the skull bone is detected. Then, all connected pixels are sequentially located and are classified into the skull bone and nonskull bone groups from the “start” to “end” point in the image [Figure 4].
Figure 4: Subtraction of skull bone detected using the “nearest-neighbor-visiting” method

Click here to view


Finally, the shapes of the left and right sides of the segmented cranium are directly compared and the related indexes are quantified. This process includes calculations of the total number of pixels, number of overlapping pixels and areas of both sides, number of nonoverlapping pixels and areas of both sides, and the CIS value. This image analysis program also computes the histogram of the cranial area and can be used to determine its composition as well [Figure 5].
Figure 5: Quantified results of the loaded image

Click here to view


Experimental validation

Plastic water pipes containing acacia gum were used as the phantom for experimental validation. We used three plastic water pipes of internal diameters (d) 72 mm (small), 84 mm (medium), and 107.5 mm (large) [Figure 6], [Figure 6]b, [Figure 6]c. These phantoms underwent CT scans, their images consisting of an outer white area resembling the skull bone, and a circular gray area resembling soft tissue [Figure 6]d, [Figure 6]e, [Figure 6]f. With PACS, the gray-colored areas were selected manually. The calculations were performed thrice for each phantom by three authors (J. M. Sun, C. T. Hsieh, and C. T. Huang). On the other hand, after the CT images were loaded into our LabVIEW-based image analysis software, two points at the upper middle and lower white circular areas were denoted to calculate the total gray areas. These calculations were also performed thrice, once by each of the three authors. The internal areas of the pipes were calculated using the formula π* (d/2)2. The percent error (%) was calculated using the following formula: actual size − estimated area/actual size × 100.
Figure 6: Plastic water pipes with three internal diameters, including 72 mm (a), 84 mm (b), and 107.5 mm (c), were scanned using computed tomography and are presented as (d-f), respectively

Click here to view


Clinical validation

We reviewed the medical records of patients who underwent surgery for CSDH at our institute from January 2009 to December 2018. The Institutional Review Board of Cathay General Hospital in Taiwan (CGH-P108069) approved our study. The inclusion criteria were as follows: (1) CSDH confirmed through brain CT without contrast enhancement, (2) presentation of only unilateral CSDH with neurological symptoms, and (3) age >20 years. The exclusion criteria were as follows: (1) concomitant occurrence of other types of traumatic brain injury; (2) history of previous neurosurgical procedures including craniectomy, craniotomy, or shunting procedures; (3) nontraumatic etiologies such as vascular abnormalities or neoplasm; and (4) missing data or images. In total, 120 patients with unilateral CSDH were reviewed. For the clinical validation, only three patients with different cranial asymmetries were investigated. We loaded the axial views of the CT images of these three patients in PACS and LabVIEW for clinical validation. PACS calculated the intracranial total area by contouring of the internal part of the skull bone. Using our LabVIEW-based image software, the total intracranial area was semi-automated calculated after denoting two points at nasion and inion [Figure 7]. The two aforementioned methods were performed thrice for each image, once by each of the three authors (J. M. Sun, C. T. Hsieh, and C. T. Huang). The CIS was also calculated.
Figure 7: Axial views of computed tomographic images of three patients with chronic subdural hematoma, who had a larger right occipital angle (a), symmetric bilateral occipital angle (b), and larger left occipital angle (c), were used to investigate cranial symmetry. The red line, also known as the midline, forms automatically upon denoting the nasion and the inion

Click here to view


Statistical analysis

Statistical analyses were performed using SPSS (version 22.0, (SPSS Inc., Chicago, IL). The data were presented as means and standard deviations. The coefficient of variation (CV) was calculated as the standard deviation divided by the mean. Between-group comparisons were performed using the independent sample Student's t-test. Differences were considered statistically significant when P < 0.05.


  Results Top


Experimental validation

The calculations using PACS and LabVIEW were performed thrice for each phantom (small, medium, and large) by three authors (J. M. Sun, C. T. Hsieh, and C. T. Huang). In total, every phantom has nine measurements from PACS and from LabVIEW. The percent error for the calculation of areas using PACS and LabVIEW versus the actual size of the phantom is summarized in [Table 1]. The actual areas of the small, medium, and large water pipes were 4072, 5542, and 9076 mm2, respectively. PACS-estimated mean areas of small, medium, and large phantoms were 3670, 5156, and 8563 mm2, respectively. LabVIEW analysis-determined mean areas of small, medium, and large phantoms were 3850, 5280, and 8725 mm2, respectively. The overall mean error of PACS and LabVIEW was 7.51% ± 1.87% and 4.68% ± 0.67%, respectively. Therefore, the results indicated that the area of the phantom estimated by the LabVIEW-based software was closer to the actual area, with significantly lower inter and intraobersever variability (P < 0.001).
Table 1: Summary of percent error of estimated areas by picture archiving and communication systems system and Laboratory Virtual Instrument Engineering Workbench-based image analysis

Click here to view


Clinical validation

CT images of three patients were used for clinical validation. The areas estimated by PACS system and LabVIEW-based image analysis are summarized in [Table 2]. Images #1 [Figure 7]a, #2 [Figure 7]b, and #3 [Figure 7]c illustrated that the CVs of the areas estimated using LabVIEW-based image analysis were 0.01%, 0.05%, and 0.01%, respectively, whereas the CVs obtained using the PACS system were 0.23%, 0.12%, and 0.16%, respectively. Other images (image #4 to image #20) showed that the CVs areas estimated using LabVIEW-based image analysis ranged from 0.01% to 0.13%, whereas the CVs obtained using the PACS system ranged from 0.05% to 0.96%. All the results thus showed that the areas estimated using LabVIEW-based semi-automated image analysis had lower variability. We also used LabVIEW analysis to assess the severity of cranial asymmetry. The CIS ratios for all images ranged from 89.2% to 99.1%, which indicated that LabVIEW analysis not only estimated the intracranial areas but also provided information on cranial symmetry for clinical application.
Table 2: Summary of estimated areas by picture archiving and communication systems system and Laboratory Virtual Instrument Engineering Workbench-based image analysis

Click here to view



  Discussion Top


A growing body of literature reveals the clinical importance of skull asymmetries.[1],[2] In an anatomic study of 200 skulls, Sarac-Hadzihalilović and Dilberović reported that the most frequent type of skull symmetry was occipital symmetry (42%), followed by left occipitopetalia (40%) and right occipitopetalia (18%).[2] Advances in computerized techniques have made computer-aided methodologies commonplace in radiological imaging for exploring asymmetric variations of pathological diseases such as stroke, hemorrhage, or neoplasms.[1] In a retrospective study of 264 health patients, Arsava et al. showed that using T1-weighted magnetic resonance images, the direction of occipitopetalia predicted the side of transverse sinus dominance.[8] Cranial asymmetry, defined by different frontal or occipital angles, has also been found to be associated with the laterality of subdural hygromas or CSDHs.[6],[7] However, all these measurements were investigated based on direct visual or simple angle measurement methods.

Although PACS is the most common software used in medical imaging to investigate areas, angles, and distances,[9] the images within PACS cannot be divided, rotated, or horizontally reflected to compare differences between the two hemispheres of the brain, which is required to accurately measure cranial symmetry. Therefore, an objective semi-automated image analysis method using MATLAB technical computing software (MathWorks, Natick, MA) with the Image Analysis Toolbox was introduced to effectively analyze cranial asymmetry.[5] The differences between bilateral crania, including asymmetry, hemispheric predominance, areas of hemispheric overlap, and CIS were calculated by simply denoting the locations of the nasion and inion. The rest of the method is automatic and calculated the area of each hemisphere and of hemispheric overlap by inverting one of the two hemispheres onto the other using the midline as the axis. However, this study only described preliminary results in measuring plagiocephaly, and the method was not applied further.[5]

In our study, we developed an image analysis software based on the LabVIEW system (National Instruments, Texas, USA) using a concept similar to that described by Zonenshayn et al.[5] Experimental validations showed that the LabVIEW-based image analysis method had significantly lower inter- and intraobserver variability compared with the results of analysis by PACS. Moreover, clinical validation showed that the CV was lower in LabVIEW-based image analysis than in PACS. Most importantly, our image analysis can semi-automatically calculate the overlapping and nonoverlapping areas of the hemispheres, with the CIS ratios reflecting the severity of cranial asymmetry accurately. We thus proved that the LabVIEW-based semi-automated image analysis is an effective and simple method of diagnosing cranial asymmetry and deformation, and it can be employed further to investigate the association between cranial asymmetry and asymmetric variation of pathological diseases.


  Conclusion Top


Our study showed that through experimental and clinical validation, the objective semi-automated image analysis based on the LabVIEW system is a convenient and effective method to investigate cranial asymmetry. It will allow us to further explore the association between cranial asymmetry and the laterality of diseases or cranial deformations.

Acknowledgment

The image analysis software was supported by LiveStrong Biomedical Technology Co., Ltd., Taipei, Taiwan.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.



 
  References Top

1.
Liu SX. Symmetry and asymmetry analysis and its implications to computer-aided diagnosis: A review of the literature. J Biomed Inform 2009;42:1056-64.  Back to cited text no. 1
    
2.
Sarac-Hadzihalilović A, Dilberović F. Study on skull asymmetry. Bosn J Basic Med Sci 2004;4:40-6.  Back to cited text no. 2
    
3.
Graham T, Adams-Huet B, Gilbert N, Witthoff K, Gregory T, Walsh M. Effects of Initial Age and Severity on Cranial Remolding Orthotic Treatment for Infants with Deformational Plagiocephaly. J Clin Med 2019;8:1097.  Back to cited text no. 3
    
4.
Ifflaender S, Rüdiger M, Konstantelos D, Lange U, Burkhardt W. Individual course of cranial symmetry and proportion in preterm infants up to 6 months of corrected age. Early Hum Dev 2014;90:511-5.  Back to cited text no. 4
    
5.
Zonenshayn M, Kronberg E, Souweidane MM. Cranial index of symmetry: An objective semiautomated measure of plagiocephaly. Technical note. J Neurosurg 2004;100:537-40.  Back to cited text no. 5
    
6.
Kim BG, Lee KS, Shim JJ, Yoon SM, Doh JW, Bae HG. What determines the laterality of the chronic subdural hematoma? J Korean Neurosurg Soc 2010;47:424-7.  Back to cited text no. 6
    
7.
Lee KS, Bae WK, Yoon SM, Doh JW, Bae HG, Yun IG. Location of the traumatic subdural hygroma: Role of gravity and cranial morphology. Brain Inj 2000;14:355-61.  Back to cited text no. 7
    
8.
Arsava EY, Arsava EM, Oguz KK, Topcuoglu MA. Occipital petalia as a predictive imaging sign for transverse sinus dominance. Neurol Res 2019;41:306-11.  Back to cited text no. 8
    
9.
Gamsu G, Perez E. Picture archiving and communication systems (PACS). J Thorac Imaging 2003;18:165-8.  Back to cited text no. 9
    


    Figures

  [Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6], [Figure 7]
 
 
    Tables

  [Table 1], [Table 2]



 

Top
 
 
  Search
 
Similar in PUBMED
   Search Pubmed for
   Search in Google Scholar for
 Related articles
Access Statistics
Email Alert *
Add to My List *
* Registration required (free)

 
  In this article
Abstract
Introduction
Materials and Me...
Results
Discussion
Conclusion
References
Article Figures
Article Tables

 Article Access Statistics
    Viewed905    
    Printed14    
    Emailed0    
    PDF Downloaded75    
    Comments [Add]    

Recommend this journal


[TAG2]
[TAG3]
[TAG4]