Simpleware Case Study: Segmentation & Modeling of Breast Tumor for 3D Printing

Introduction

A cancer diagnosis is always a difficult reality to face. For most patients, medical environments and terminology are foreign and difficult to process without the stress induced by the “c-word”. Moreover, treatment for most pathologies is highly dependent on patient compliance with prescribed regimen. This requires a trusting partnership between the patient and treating physician where effective communication and dependable follow-through takes place. Patient education is a necessary step in building accurate awareness of the challenges being faced, how the treatment plans address the challenges, and the distinct roles played by patient and physicians. 

Physical 3D printed models have been shown to be highly effective education tools and enable better visualization of the anatomy of the pathology. This case study applies the segmentation tools within Simpleware software to easily generate 3D models of the breast and tumor of a breast cancer patient in order to build common ground through patient education by presentation of 3D printed models. This case also demonstrated the added benefit of enhancing the physicians’ spatial conceptualization of the tumor location and shape.

The MRI dataset was received in DICOM format and imported into Simpleware ScanIP for segmentation. After adjusting the Window/Level settings for optimal viewing, the Threshold tool was used to set grayscale value ranges for masking the breast tissue generally and the tumor specifically. The Crop tool was used to isolate the left breast as the region of interest. The Mask Flood Fill tool allowed for the clean segmentation of the tumor and breast tissue by using it on the tumor mask and the breast mask respectively. 

This fill step ensured only the connected voxels remained in the mask data. The proper coverage of the masks was further optimized using the Morphological Close and Cavity Fill tools to efficiently fill gaps in mask continuity. The Paint tool allowed for user guided judgement to put the finishing touches on the masks to ensure voxels were either masked or unmasked as needed. After the masks were complete for the breast and tumor regions in the images, the Resample tool rendered isotropic voxels. Then the masks were smoothed by first using Dilate then Recursive Gaussian Smoothing which generated a clean, smooth surface. Finally, the masks were exported as STL models for further processing prior to printing.

Model Generation

Segmentation results were exported as two .stl surface models: breast and tumor. These were then imported into Autodesk Fusion 360® for further modification before printing. The breast model was hollowed such that the printed part would represent the general external shape but without internal fill, so the tumor model could be both placed and visualized inside the breast. A post was also designed to connect the tumor and breast models after printing. The post also ensures that the tumor will be located in the proper position after printing for accurate representation. The breast model was also divided into two halves to make visualization easier.

3D Printing

The models were imported into Stratasys GrabCAD™ for arrangement, processing, and sending to the 3D printer. The models were printed at Forsyth Technical Community College on a Stratasys J750™ with Vero material which allows for multiple colors from clear to black.

The three parts of the model fit together easily. The two halves of the breast were easy to align. Design features could easily be added in future iterations to hold them together if needed. Placement of the tumor model with integrated post into the designed post-hole in the breast model made registration of one model to the other simple and quick. By placing the post in the hole, the tumor was immediately in its appropriate location relative to the breast model. Size and location of the tumor within the breast model are the most valuable implications of the printed model, therefore having a model of the tumor and breast to scale and a simple spatial orientation mechanism makes this design immediately useful.

Physician Feedback

Collaborating physicians were thoroughly impressed with the outcome and saw several benefits beyond the scope of this project. They confirmed its utility in aiding surgeons in planning and communicating about approaches to resection and other treatment modalities. The potential to increase patient comprehension of the implications of the pathology also excited the physicians. Improved patient understanding through visualization will help the conversation between doctor and patient about needs and implications of various treatment options. 

For instance, visualizing the tumor size and location can facilitate the conversation regarding the decision between lumpectomy to remove the tumor or mastectomy to remove the entire breast. This conversation is sensitive, and visualization can be a helpful tool. Furthermore, the visual aid may help promote patient compliance and morale through better awareness of the goal for treatment regimens.

Conclusions

This approach to visualizing human anatomy has far reaching implications. Here, we discuss the impact on patient education, but there is more that can be done with respect to physician guidance. Physician education can also be improved as models of the pathology and surrounding tissues can be printed with materials that mimic their appearance and tactile properties. The use of multi-model, multi-material printing to produce patient specific surgical models will enhance the physician’s ability to treat complex problems with the potential to reduce surgical time, reduce collateral damage to surrounding tissue, and improve efficacy of treatments by practicing directly on a model of that patient’s pathology. 

Another potential use for efficient formation of segmented models is that of patient-specific bioprinting. Modeling both the healthy and diseased tissue can lead to data useful for bioprinting replacement tissue that can be later implanted to replace or regenerate the affected area. Instead of slicing the model and printing it with plastics, the bioprinting process uses biodegradable plastics and hydrogels to print multiple cell types in organized patterns designed to regenerate specific functional tissues. The implant could be a tissue like liver or a composite set of tissues like muscle, tendon, and bone.

3D Printed anatomical models are beginning to have a significant impact on medicine. The first step for patient specific model printing is segmentation of patient imaging data sets. Simpleware ScanIP makes that process simple and efficient as printable models can be quickly produced from segmented portions of the imaging data. The model of the breast along with a tumor being treated allowed the patient to better understand the diagnosis and subsequent treatment options being proposed by the physicians. The models were also useful for the physicians to better visualize the anatomy of the pathology.