This prospective study enrolled patients receiving conservative treatment for musculoskeletal complaints between November 30, 2019, and April 25, 2021. Inclusion criteria were (1) age ≥ 18 years; (2) recent history of trauma; (3) diagnosis of fracture on digital radiography (DR) or suspected fracture; and (4) clinical indication for SD-CT. Exclusion criteria were (1) metal implant; (2) history of tumors; or (3) history of arthritis or bone metabolic disease.

Included patients underwent clinically indicated SD-CT followed by an ULD-CT at an interval of 1–2 weeks. 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 [13].

The protocol for this study was approved by the Ethics Committee of  Guangdong Provincal Hospital of Traditional Chinese Medicine (BF2019-030-01). All patients provided written informed consent for the acquisition of a ULD-CT after a SD-CT.

Scans were performed using a Canon 320-detector-row CT scanner (Aquilion One Vision; Canon Medical Systems, Otawara, Japan). For the shoulder, knee, ankle, and wrist, SD-CT scanning parameters were 120 kV tube voltage and 150, 120, 120, and 50 mA tube current, respectively; ULD-CT scanning parameters were 80 kV tube voltage and 52, 11, 11, and 4 mA tube current, respectively; scan range was 160 mm, 140 mm, 140 mm, and 100 mm, respectively. Scan slice thickness was 0.5–1 mm. CTDIvol (mGy) and DLP (mGy*cm) were automatically implemented for all CT-protocols by the scanner software.

Effective dose (ED = DLP*k) for each patient was calculated by multiplying DLP by k (a conversion coefficient): shoulder k = 0.0113 (SD-CT) k = 0.0091 (ULD-CT); knee k = 0.0004 (SD-CT and ULD-CT); ankle and wrist k = 0.0002 (SD-CT and ULD-CT) [12]

Post-processing was performed on a dedicated workstation (VitreaFX3.0). Image reconstruction involved multiplanar reformatting (MPR), volume rendering (VR), and maximum intensity projection (MIP).

Two senior clinicians with 10–13 years of experience in musculoskeletal diseases independently reviewed each image to characterize each fracture as displaced or non-displaced. Displaced fractures were defined as having a fracture line > 2 mm wide and/or > 1 mm displacement of the bone cortex. Non-displaced fractures were defined as having no angulation or shortening, a fracture line < 2 mm wide, and/or < 1 mm displacement of the bone cortex [14,15,16]. Avulsion fractures caused by a sudden and violent pull of a muscle or ligament were characterized as displaced or non-displaced fractures when bone fragment displacement was > 5 mm or < 5 mm, respectively [16]. Each clinician reviewed each image twice at an interval of > 6 weeks. Disagreements about image interpretation were resolved through discussion and consensus.

A final diagnosis was made based on the CT/DR review within 1–3 months based on the presence of a callus at the fracture end, dysplasia, and an old fracture without a callus [8, 16].

One experienced radiologist evaluated objective CT image quality metrics. A region of interest (ROI) (70 mm2) was placed within the muscles around the joints. Mean/standard deviation CT values of muscle (CTm) were determined from three measurements. A ROI (8 mm2) was placed on the thickest region of the cross section of the cortical shell of the bones of the joint. Mean/standard deviation CT values of bone (CTb) were determined from three measurements. CT values of joint cortical bone (CTc) were calculated as: CTb-CTm. Noise was calculated as mean CTm standard deviation. Signal-to-noise ratio (SNR) was calculated as: mean CTm/mean CTm standard deviation. Contrast-to-noise ratio (CNR) was calculated as (mean CTc–mean CTm) /mean CTm standard deviation [16].

Two experienced radiologists and two orthopedic physicians evaluated subjective CT image quality and the impact of subjective CT image quality on clinical decision-making on a 5-point Likert-type scale (Table 1).

Statistical analyses were conducted using IBM SPSS Statistics, v26.0 (IBM Corp., Armonk, NY, USA). CTDIvol, DLP, ED, CTc, SNR, and CNR for SD-CT and ULD-CT were compared using analysis of variance (ANOVA), or Tamhane's T2 test for data with unequal variances. Subjective CT image quality and the impact of subjective CT image quality on clinical decision-making for SD-CT and ULD-CT were compared with the rank sum test. The consistency of the two radiologists on the 5-point Likert-type scale evaluating subjective CT image quality and the impact of subjective CT image quality on clinical decision-making was assessed using the intraclass correlation coefficient (ICC), where < 0.40 = poor consistency; 0.41–0.60 = moderately consistent; 0.61–0.80 = good consistency; 0.81–1.00 = perfect consistency. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of SD-CT and ULD-CT for the diagnosis of non-displaced fractures of the shoulder, knee, ankle, and wrist were calculated. Observer performance for ULD-CT and SD-CT was estimated by calculating the area under the Receiver Operating Characteristic (ROC) curve (Az).