3D-Printed Phantom: Lower-Dose CT for Infant Head Imaging

CT head imaging may be necessary for very young patients, but children are more sensitive to ionising radiation than adults. The challenge is to tailor scanning settings to a child’s size while still producing images that provide enough information for diagnosis.

A 2023 paper by Adi Pandžić, PhD, in Diagnostics describes how a 3D-printed anthropomorphic infant head phantom was made and used to optimise a CT protocol.

Key findings

• The phantom was made from a 3D model based on DICOM images, using SLA 3D printing, plaster and epoxy resin.
• The optimised protocol used a tube voltage of 100 kV, a tube current of 170 mA and a rotation time of 0.75 s.
• Mean CTDIvol was 27 mGy for the unoptimised protocol and 21 mGy for the optimised protocol; the difference was statistically significant (p = 0.021).
• The subjective image-quality assessment found no statistically significant difference between the protocols (p = 0.379).
• Image quality under the optimised protocol was assessed as acceptable, but the phantom does not reproduce every structure of the human brain.

Why paediatric CT settings need attention

CT uses X-rays to show internal anatomy. In paediatric radiology, an adult protocol is not automatically suitable: children have smaller bodies and anatomical structures, and their tissues are more sensitive to radiation. Their longer life expectancy also makes careful optimisation important. In practice, this means keeping exposure as low as reasonably achievable while maintaining the image quality needed for the medical purpose.

In the clinical setting described in the paper, one head CT protocol was being used for children across a wide age range, from 0 to 16 years. The study focused on infants aged 0–1 year as an initial step towards age-based optimisation. The paper explains that a model was needed to test scanning settings and measure image properties, with a shape and radiological response more appropriate for an infant head than a commercial adult phantom.

Building a 3D-printed infant head phantom

A phantom is a physical model that imitates selected properties of the body. In CT, phantoms can be used to assess how different materials appear and to check imaging settings. The study used a commercially available ATOM Max adult head phantom as a reference. Its materials have X-ray attenuation properties similar to normal tissue, but its size and homogeneous representation of brain tissue limit its suitability for assessing paediatric anatomy.

The infant model was developed from CT image data stored in DICOM format. The paper reports that 3D Slicer was used to set a threshold for bone, segment the bone tissue and create a three-dimensional model. Information that could identify the patient was removed from the DICOM data.

The model was exported as an STL file and prepared for printing in PreForm. The study used stereolithography, or SLA, and Formlabs Grey V4 material to print the skull structure and casting mould. The mould was then filled with plaster and epoxy resin and allowed to become rigid before CT imaging. Plaster was selected to simulate bone; the CT values of the plaster used corresponded to values measured in children’s skull bones.

The paper notes that radiographic properties of 3D-printing materials, including CT-number characteristics and density uniformity, had been examined previously. This study builds on that materials work and describes how the phantom was applied to CT protocol optimisation.

What was measured and how settings were selected

Testing took place on a 16-slice Toshiba Astelion CT scanner. The adult head protocol used as a reference had a tube voltage of 120 kV, a tube current of 200 mA and a rotation time of 750 ms. Tube voltage is expressed in kilovolts, while tube current and scan time together determine the amount of radiation used during an exposure.

The study measured CT numbers and image noise in the phantoms. A CT number, expressed in Hounsfield units (HU), describes how strongly a material attenuates X-rays compared with water. Image noise is random variation in pixel values and affects the appearance of a scan. Measurements were taken in a circular region of interest with an area of 500 mm². The study also evaluated CTDIvol, a volume CT dose index, and PKL,CT, the dose-length product.

Based on the phantom measurements, the study selected an infant protocol with a tube voltage of 100 kV, a tube current of 170 mA and a rotation time of 0.75 s. The clinical comparison included 12 examinations performed with the unoptimised protocol and 30 with the optimised protocol. A radiologist subjectively assessed five image-quality criteria, including noise and overall acceptability, grey–white matter differentiation, subarachnoid-space visibility, visualisation of the posterior cranial fossa, and streak artefacts.

Results: dose and image quality

Mean CTDIvol in the unoptimised group was 27 mGy, with a standard deviation of 14 mGy. In the optimised group, the mean was 21 mGy, with a standard deviation of 0 mGy. The difference in CTDIvol was statistically significant (p = 0.021). The paper also reports that dose values were more consistent under the optimised protocol. The wider spread in the unoptimised group may have been related to arbitrary selection of scanning settings.

Mean PKL,CT was 432 mGycm for the unoptimised protocol and 365 mGycm for the optimised protocol. These values also varied with scan length. The subjective assessment found no statistically significant difference in image quality between the groups (p = 0.379). Mean grades for all five image-quality criteria were above 2, which the paper describes as acceptable or good image quality.

What the findings mean in practice

For radiology teams, this study illustrates how phantom measurements can support age-specific CT protocol adjustments. A phantom allows settings and image properties to be examined without relying only on patient scans. The results apply to the scanner, materials and clinical protocol evaluated at the Sarajevo University Clinical Center; they should not be taken as a universal set of settings for every CT system or clinic.

For engineers and researchers, the paper documents a workflow spanning DICOM data processing, anatomical segmentation, STL preparation, SLA printing and the use of different materials to represent bone and soft tissue. It is an example of additive manufacturing in medical imaging and of tailoring a phantom to a particular anatomy. The paper brings together mechanical engineering, medical physics and radiology.

Limitations and next steps

The phantom has limited structural complexity, so contrast analysis could not be carried out. The paper also reports that the epoxy resin does not have the same HU values as human brain matter. For this reason, the phantom could not be used in radiotherapy. In addition, the adult commercial phantom represents brain tissue homogeneously and does not distinguish grey matter, white matter or cerebrospinal fluid in the ventricles.

These limitations constrain what can be concluded from phantom measurements. The findings support using this model to optimise CT parameters in the specific context studied, but they do not establish that the same materials or settings can replace clinical evaluation elsewhere. Developing more realistic physical phantoms could allow more detailed assessment of anatomy and image contrast.

Conclusion

The study demonstrates that SLA 3D printing and commercially available materials can be used to produce an anthropomorphic infant head phantom. In this specific evaluation, the optimised CT protocol had a lower mean CTDIvol than the unoptimised protocol, while subjective assessment found no significant difference in image quality. The phantom therefore served as a tool for protocol optimisation, with clear limitations in how it represents brain tissue and anatomical complexity.

Frequently asked questions

Author: Adi Pandžić, PhD in Mechanical Engineering

This article summarises a published research paper; please refer to and cite the original paper. About the author

← All articles