This approach utilizes of one of the most affordable 3D-printing methods, FDM, to prepare complex models. Models using previous “standard” techniques had estimated prices of approximately $1000 for a liver model [10] and $500 for a kidney [11] model. Thanks to cost-effective approach, it seems possible to implement this solution in most hospitals worldwide. The technique can be executed by physicians after a relatively short training period.

3D-printed models are used in surgery to plan and prepare for extensive procedures, as a intraoperative guidance as well as an anatomy and pathology training tool [1]. Not only do models help surgeons, but they can also be used in many out-of-surgery applications, including education of students and young doctors in subjects such as anatomy, pathology, and surgery [12, 13] and in patient education [14]. All these make this low-cost method described in this paper feasible and justified among healthcare providers to implement 3D printing in clinical situations, which was previously available only to a limited number of hospitals.

Physical models also may be found more beneficial than 3D-rendered images. Virtual models do not represent structures in 1:1 scale, they also tend to be difficult to interpret in cases with complex anatomy, especially when evaluating the course of blood vessel branches. Although, there are only a few studies on this topic [5].

To date, PolyJet/MultiJet (photopolymer-based) or, more rarely, stereolithography (SLA) technologies have primarily been used to prepare similar models for pre-surgical preparation. However, PolyJet printers are industrial-grade, and their price varies from $6000 (low-end printers) to over $200,000 [8]. The high-end printers have been used to create pre-surgical anatomical models [11, 15]. Desktop 3D FDM printers are generally available starting from few hundred dollars and range up to $5000 for new-generation, relatively sophisticated FDM printers [16].

Accuracy of our model is affected by several factors. First, the resolution of the CT images most likely has the largest effect on the outcome. Some studies report that when a slice thickness of 2.5 mm is used, as in the present case, the liver volume may be underestimated [17]. However, semi-automated segmentation, as was presented in this paper, is reported to slightly overestimate the volume [18]. Computer processing and PLA finishing may have some effect on the volume, as some studies report 3D smoothing contributes to model shrinking [19], similar to coating PLA surfaces with thin layers of resin. Most of the described components mainly affect the parenchyma rather than the vessels or the tumor. Accurate calculations and dimension comparisons should be performed under conditions that allow for the measurement of the actual size of the liver. The ideal scenario for these measurements would be a transplantation surgery.

The accuracy of the model can be improved. For example, different algorithms can be utilized for segmentation, and CT scans with a lower thickness should be tested.

This technique requires creation of multiple parts of one object (for example, the liver parenchyma) to assemble the model, cast the silicone inside, and demount external parts afterward. Printing an entire model from one STL object would result in printouts filled with support material that would be impossible to remove. Additionally, it would not be possible to remove external parts to obtain the liver-shaped silicone parenchyma. This “multipart approach” also avoids limitations of the small printing field. Many 3D printers have printing field sizes smaller than 200\(\,\times \,\)200 mm, which is usually smaller than a full-sized liver. The full size models may be printed due to division. Many previous researchers describe the inevitability of scaling down models due to cost or printer limitations [10]. Moreover, smaller parts are usually easier to print (fewer artifacts and simpler removal of support material).

To date, few reports have described methods to correctly arrange multipart, complex structures, and some have reported the use of positioning joints and pins [20]. Our solution, using Boolean-based parts, is in our opinion accurate and does not disrupt the original structure.

Compared to significantly more expensive methods, our technique is slower (print time of 60–100 h compared to 36–40 h [6, 10]). However, the printing time can be reduced by half or more with the use of multiple printers, which still would be far more cost-effective. In addition, PLA printouts are more fragile than PolyJet-based printouts, especially during postprocessing. In addition, the assembly must be very accurate due to the risk of silicone leakage and the loss of model precision.

Liver malignancies surgery is not the only application of this method or of 3D-printed models. The relatively long printing time may limit the use of this technique in emergency surgeries; however, elective procedures, which represent the majority of hepatic resections [21], can greatly benefit from this 3D printing method.