Feasibility and fidelity of 3D printed ASD models based on
3D-TEE images
Feasibility and fidelity are significant prerequisites and guarantees
for the medical application of ultrasound-derived 3D printed models.
However, steps involved in creating a 3D printed model were meticulous
and complex. Errors are easily introduced in the sophisticated workflow
of image acquisition, image post-processing and 3D model manufacture,
which may decrease the fidelity of 3D models. Therefore, the ability to
highly retain the information of original 3D-TEE images is the key for
the 3D models to combine the advantages of 3D-TEE and 3D printing
technology. Quantitative assessment of fidelity is of great significance
for subsequent application in ultrasound-derived 3D
printing11.
In this preliminary study, all the 3D-TEE ASD images were successfully
post-processed and converted to solid 3D ASD models, demonstrating the
feasibility of echocardiography used as the data source for the 3D
printing of models, which requires a trained echocardiographer, a 3D
echocardiography system, a post-processing software for data conversion,
segmentation, modeling and a 3D printer 19. The first
condition for successfully obtaining 3D ASD models is to acquire the
3D-TEE images and converted to a DICOM format that can be accessed by
offline post-processing software. The ultrasound systems used in this
study are capable of exporting 3D data in the DICOM format, in which the
echocardiographic images can dynamically display the real-time changes
of ASD in each phase during the cardiac cycle. By replaying the dynamic
characteristics of the observed images, the key frames at the
isovolumetric diastole are selected for image post-processing. In
addition, whether ultrasound-derived 3D printing is feasible depends on
post-processing software. Mimics software applied in this study is one
of the available post-processing software that have a wide selection of
manual and automatic segmentation tools and capable of correcting mesh
files,which allowing to obtain patient-specific modelling and
meshes19. The last basis for 3D printing is 3D
printer, the FDM 3D printer used in our study can generate instructions
and build the 3D ASD model layer by layer completely.
ASD occlusion is considered to be the first-line therapy for patients
with morphologically suitable secundum ASD20. For
achieving successful closure of ASD, accurate determination of the
device size is one of the critical steps in the procedure. The maximum
diameter is considered to be the most important reference parameters in
choosing the appropriate size of an ASD occluder with central
waist21. However, the shape of the ASD is oval or
irregular rather than round occasionally, self-expansion properties of
the ASO may cause deformation of the ASD. Noticeable changes may
generate in the maximum diameter and the minimum diameter of ASD
simultaneously and the degree of change depends on the shape of ASD.
Studies have pointed out that stretching of the oval ASD by the device
will probably occur by preference at the minimum dimension and make the
ASD more circular22. A single maximal diameter
measurement might overestimate the device size in patients with an oval,
large ASD and cause the selection of devices
oversized23. Considering that the maximum diameter and
the minimum diameter will change with the implantation of the occluder,
the parameters that do not change with deformation should also be
analyzed, especially the circumference. The circumference may be a
meaningful reference to estimate device13.
Therefore, parameters measured from 3D printed models in this
preliminary study have reference significance for ASD closure, such as
the maximal diameter, the minimal diameter, circular index and
circumference. It showed no significant differences among the three
methods. What’s more, the Bland-Altman scatter plot demonstrates that
the ASD size parameters were concordant well among 3D printed
models,2D-TEE images and 3D-TEE images. Moreover, the absolute
difference was small among the three methods as well. All these results
indicate that the 3D printed ASD models precisely preserve the original
information of the ultrasound images and the fidelity of
ultrasound-derived 3D ASD models can be guaranteed if meticulous care is
taken in the workflow. 3D printed models have the potential to reflect
the ASD anatomy accurately.
This preliminary study is focused on process-oriented outcomes to
evaluate whether tiny errors were introduced in every step. The high
fidelity of the ASD 3D printed models mainly depend on tight error
control during data acquisition, image post-processing and printing
described herein 24. Firstly, the acquisition of
optimized 3D-TEE ASD images is the basis for accurate 3D printed models.
Newer echocardiography transducers, coupled with the continuous advanced
ultrasound hardware and software equipment makes it possible to obtain
3D echocardiographic images with high resolution. The grayscale
variations within the structure of interest were minimized and the
contrast between ASD and surrounding area were maximized. The desired
ASD 3D echocardiographic data set we finally obtained have a completely
dark blood pool and completely bright myocardium with minimal noise and
motion blurring. Secondly, the keys for a precision post-processing are
to recognize fine details of ASD and adjust the thresholding to the
optimal. During post-processed, the upper and lower limitation of
thresholding was altered according to the endocardium meticulously.
There were no cases of missing ASD structures or excessive image noise
caused by improper threshold settings and manual segmentation errors.
The balance between signal and noise of the 3D ASD digital models were
maintained and the important anatomic details of ASD were saved well.
Thirdly, the fidelity of the 3D printed models also benefited from the
high resolution of the printer and its advanced control software. FDM
printers can provide high resolution ranging from 0.1 to 0.2mm after
considerable post-processing, tiny errors in the process of 3D model
manufacture can be avoided effectively24. In this
study, quantitative comparisons of the results with the original data
set at every stage of processing were taken carefully. It can be
confirmed that the 3D printed model can highly retain and reproduce the
original 3D-TEE image information.