This study was approved by the institutional research ethics committee of the Guangdong Provincial Hospital of Traditional Chinese Medicine (BF2019-030-01). Written informed consent of patients was obtained. Between February 2019 and August 2020, 36 patients (21 men and 15 women) were treated at the Zhuhai Provincial Hospital of Guangdong Hospital of the Traditional Chinese Medicine Hospital. An average of two knee joint scans were performed per patient. These included 20 patients who were undergoing conservative treatment of knee fractures (CT re-examination was required for non-operated fracture patients within 2 weeks to evaluate any increase in the degree of fracture displacement necessitating surgery), 2 patients with a history of repeat injury after the first CT scan, and 14 patients who underwent cruciate ligament reconstruction (one CT scan each before and after surgery). The first CT scan was performed using standard-dose protocol, while the second CT scan used ultra-low-dose protocol; the mean interval between the two scans was 8.0 ± 2.1 days. The mean age of patients was 43.8 ± 15.3 years (range 19–82). Patients aged ≥ 18 years were included. Patients aged < 18 years and pregnant women were excluded.

All CT scans were performed using Canon 320-slice dynamic CT (Canon Aquilion One, Japan). According to the previous research results, the CT scan parameters were as follows: FOV: 20 mm; reconstruction layer thickness: 0.5 mm; spacing: 0.5 mm. In conventional CT scan (according to the textbook and parameters of Chinese medical imaging technology CT imaging technology jointly developed by Canon Engineers), CT scan parameters were as follows: tube voltage 120 kV; tube current 100 mAs. For ultra-low-dose CT, the scanning parameters were as follows: tube voltage 80 kV, tube current 11 mAs. The scanning conditions for each group are shown in Table 1.

The original CT data were reconstructed iteratively to obtain the data package. The layer thickness and layer spacing of the data package were 0.5 mm. The data package was imported in the reconstruction software Mimics research 21.0 and the slice software (Simplify3D V4.0). According to the reconstruction data, the skeleton model using the melting deposition method was printed using Tianwei coliDo 3.0 3D printer. The printing material was diameter 1.75 mm white polylactic acid.

Computed Tomography Dose Index volume (CTDIvol) (mGy) and Dose length product (DLP) (mGy * cm) were automatically indicated by the scanner software for all CT-protocols. In order to obtain the effective dose (ED = DLP * k), DLP for each patient was multiplied by K (a conversion coefficient). For the extremity scans, k = 0.0002 has been used for extremity scans [17].

Vernier calipers were used to measure the following indicators on the 3D printed model of knee joint: anterior and posterior diameter and the left and right diameter of the internal and external condyles of the femur (APDLF: Anterior and posterior diameter of lateral femoral condyle; APDFC: Anterior and posterior diameter of the femoral medial condyle; LRDMLFC: Left and right diameters of medial and lateral femoral condyles). Left and right diameters and anteroposterior diameters of tibial plateau (LRTP: Left and right diameters of tibial plateau; APTP: anterior and posterior diameters of tibial plateau). The upper and lower meridian and left and right diameters of the patella (ULP: Upper and lower meridian of the patella, LRDP: Left and right diameters of the patella). The anteroposterior diameter, left and right diameter of the femur, the anteroposterior diameter of the tibial plateau, the left and right diameter, the distance from the intercondylar ridge to the tibial tuberosity, the lower femur angle, and the upper tibial angle were measured on the CT images of the knee joint (Canon Workstation, Canon Aquilion One, Japan). The above parameters play an important role in the selection of the internal fixation model. Three measurements were taken, and the average value was used for analysis.

Two senior radiologists with at least 7 years of experience in diagnostic radiology participated in this evaluation. They had not participated in the clinical optimization of the previous hybrid system and were blinded to the patient’s clinical information. Both radiologists independently reviewed the collection of images in a random order using a dedicated workstation. Individual optimization was also allowed for the window, including the subjective image quality and detectability of structures. Referring to scholar CT scan for image quality assessment [18, 19], the image quality of anatomical structures and image noise was scored on a 3-point scale defined as follows: Score 1–Distortion of spatial resolution or contrast resolution or impossible edge definition due to high image noise; image quality is poor and unevaluable; image noise is very high with reading restrictions, and a diagnosis is difficult or even impossible. Score 2–Adequate image quality, affected by image noise, or some spatial resolution or contrast resolution distortion; but edge definitions are fully present; images can be read without distraction, enough image noise is present, not distracting reading, but still allowing a correct diagnosis. Score 3–Good image quality, with only minimal image noise, or minimal spatial resolution or contrast resolution distortion; image noise has no effect on fracture diagnosis.

Two senior orthopedic attending doctors used the 3-point evaluation method to assess the definition of the 3D printed model and the guidance of the operation. 3 points: the 3D printed model has a smooth surface, which has no adverse influence on the design of the operation plan, preoperative operational practice, and intraoperative auxiliary operation [20]; 2 points: slightly rough surface of 3DP model, which has no adverse influence on the design of the surgical scheme, preoperative simulation of the operation, and the intraoperative auxiliary operation; 1 point: rough surface of the 3DP model, which has a great influence on the design of the surgical scheme, preoperative simulation of the operation, and intraoperative auxiliary operation [20].

Thirty-six cases of knee joint patients with standard-dose and ultra-low-dose data packets were sent to AK Medical Group, and the mincs system was used to simulate total knee replacement to select surgical methods and internal fixation models.

Statistical analyses were conducted using SPSS software, version 27.0 (IBM Corp, Armonk, NY, USA). Data normality was assessed using the T test. Differences with respect to CTDIvol, ED, and objective indicators were assessed using the t test. Image quality scores and 3DP model scores were analyzed using the rank sum test. The Kappa test was used to estimate the consistency between the evaluation of 3D printing model by 2 doctors (Kappa value < 0.04: slightly consistent; 0.41–0.60: moderately consistent; 0.61–0.80: highly consistent; 0.81–1.00: almost completely consistent). p values < 0.05 were considered indicative of statistically significant difference.