Three-Dimensionally Printed, On-Demand, Patient-Specific Breast Models to Support Mastectomy Gross Examination: A...

Accurate gross examination of mastectomy specimens relies on accurate correlation between preoperative imaging and specimen anatomy to guide sectioning, orientation, and margin-oriented sampling. Radiologic-pathologic spatial discordance remains a common source of variability, particularly in multifocal tumors and in neoadjuvant-treated specimens with poorly defined tumor beds. Three-dimensional (3D) printing enables the production of patient-specific physical models that may assist in tumor localization and spatial understanding during macroscopic assessment. In this proof-of-concept study, two representative breast magnetic resonance imaging (MRI) examinations were selected from the DUKE-BREAST-CANCER-MRI dataset of The Cancer Imaging Archive (TCIA). The two cases represented unifocal and multifocal breast cancer, respectively. Segmentation of the breast contour and lesions identified as tumors was performed in both cases, with additional segmentation of a possible enlarged axillary lymph node in the second case. Anatomical orientation labels, a nipple landmark, and empty rectangular labels for each case number were incorporated into both models. The projection of the lesions onto the anterior external breast contour was encoded using protruding lines to indicate their relative surface extent. Each model was printed in separate parts, sectioned along the sagittal plane, to indicate the relative position of the tumor(s) in sagittal cross-section. All models were printed using polylactic acid (PLA) filament on a Bambu Lab A1 3D printer (Bambu Lab, Shenzhen, People's Republic of China). Printing time and material use were recorded at the maximal print speed preset of 166%. A Pareto analysis, on printer-estimated total print time and slicer-estimated total material use, was performed on the two-tumor model, with identification of the respective Pareto knees. An additional third model was produced from an MRI examination showing a lesion with associated breast skin thickening and enhancement. This model was printed in two colors to aid the identification of the extent of breast skin with MRI abnormalities. For all three models, a cost estimation was performed, incorporating material use, electricity consumption, and final print weight after support removal for the estimation of incineration cost. In conclusion, MRI-derived, patient-specific 3D-printed breast models may be produced in modest time and at low cost within a pathology-oriented workflow. They may be used not only for spatial orientation but also for educational and research purposes. This proof-of-concept study supports further evaluation of such models as spatial orientation aids for gross sectioning, tumor bed localization, and margin-oriented sampling, particularly in cases with radiologic-pathologic discordance and in neoadjuvant-treated specimens. Future clinical decision support applications would require the development of a validated workflow for model production.