Geometric plane alignment as a methodological requirement for accurate angular measurement in three-dimensional craniofacial imaging Authors: Mert Nahir Published: 2026 ABSTRACT Accurate angular measurements derived from three-dimensional (3D) radiological imaging are essential for morphometric analysis, surgical planning, and implant design. However, measurements performed on standard sagittal, coronal, and transverse sections may not reflect true spatial geometry due to projection-related distortions. This study aimed to develop and validate a reoriented plane protocol to improve the accuracy and reproducibility of angular measurements in 3D imaging.Five anatomically accurate plastic skull models were used to define four craniofacial angles (∠Co–Id–Co, ∠Go–Gn–Go, ∠Co–Go–Gn, ∠N–Rza–Pr), and all measurements were independently performed by three observers. Direct goniometric measurements served as the reference standard. CT scans (0.625 mm slice thickness) were analyzed using standard sectional planes and a reoriented plane approach aligning the measurement plane with the true spatial configuration of landmarks. Intraobserver and interobserver reliability were assessed using intraclass correlation coefficients (ICC), and agreement was evaluated with Bland–Altman analyses.Goniometric measurements showed mean values of 55.8°±4.2°, 72.9°±4.9°, 113.5°±6.5°, and 36.5°±5.7° for the respective angles. Reoriented plane measurements produced nearly identical results with minimal bias (e.g., 0.002° for ∠Co–Go–Gn) and narrow limits of agreement. In contrast, standard sectional measurements demonstrated substantial deviations, with mean biases reaching − 36.47° in coronal sections. The reoriented plane method showed excellent intraobserver and interobserver reliability (ICC up to 0.950, p < 0.001).Aligning the measurement plane with the true landmark-defined plane significantly improves angular measurement accuracy and reproducibility in 3D radiological datasets, offering a robust methodological framework for clinical and research applications.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12903-026-08533-x. Background Three-dimensional (3D) radiological imaging is now fundamental to modern medicine, supporting diagnosis, surgical planning, and research activities [1–3]. These techniques improve diagnostic accuracy by delivering high-resolution visualization of anatomical structures and pathological conditions, thereby facilitating comprehensive preoperative evaluation [4, 5]. They are also widely used in epidemiological and morphometric research [6, 7]. Specialties including orthopedics, neurology, cardiology, and surgery rely extensively on these imaging techniques in both clinical and research settings [4, 5, 8]. Volumetric and linear measurements derived from 3D radiological images provide essential data for biomarker identification, surgical planning (orthopedic, orthognathic, and reconstructive), implant design, and deformity analysis. Thus, beyond diagnostic and therapeutic applications, 3D imaging represents a fundamental resource for scientific research. However, although measurement from 3D datasets may appear straightforward, lack of standardized protocols can lead to inaccurate results. If anatomical structures are not assessed within the correct reference planes, measurements may deviate substantially from true anatomical values. In particular, angular measurements performed without proper plane alignment may produce over- or underestimated values (Figs. 2, B3), potentially resulting in clinically significant consequences such as flawed surgical planning, inaccurate morphometric analysis, or improper implant fabrication. Therefore, establishing standardized protocols to improve the accuracy and reproducibility of 3D radiological measurements is essential. The present study aimed to develop a protocol for precise and reliable angular measurement in three-dimensional imaging. Materials and methods Model construction and study design Anatomically accurate plastic human skull models were used to ensure standardization and eliminate variability associated with dry bone specimens, such as surface wear or deformation. The models (Somso®, Coburg, Germany), widely used in medical education, are recognized for their dimensional fidelity to the human skull and are suitable for methodological studies evaluating the accuracy of angular measurements derived from three-dimensional (3D) radiological images. Four angular parameters were defined for measurement: ∠Co–Id–Co: Condylion – Infradentale – Condylion.∠Go–Gn–Go: Gonion – Gnathion – Gonion.∠Co–Go–Gn: Condylion – Gonion – Gnathion.∠N–Rza–Pr: Nasion – Root of the zygomatic arch – Prosthion. ∠Co–Id–Co: Condylion – Infradentale – Condylion. ∠Co–Id–Co: Condylion – Infradentale – Condylion. ∠Go–Gn–Go: Gonion – Gnathion – Gonion. ∠Go–Gn–Go: Gonion – Gnathion – Gonion. ∠Co–Go–Gn: Condylion – Gonion – Gnathion. ∠Co–Go–Gn: Condylion – Gonion – Gnathion. ∠N–Rza–Pr: Nasion – Root of the zygomatic arch – Prosthion. ∠N–Rza–Pr: Nasion – Root of the zygomatic arch – Prosthion. The anatomical landmarks used to define the angles are presented in Fig. 1. Fig. 1Anatomical measurement landmarks forming the angles Anatomical measurement landmarks forming the angles Step 1: reference parameters Angular measurements were obtained directly from the models by three observers using a goniometric ruler. Each angle was measured three times, and mean values were calculated and recorded. These measurements served as the reference standard for comparison. A custom three-dimensional goniometric ruler was designed and produced using a 3D printer to enable accurate measurement on complex craniofacial structures (Fig. 2). Conventional goniometers were unsuitable due to the skull’s three-dimensional geometry. Fig. 2Measurement of the ∠N–Rza–Pr angle on the model using a customized goniometer Measurement of the ∠N–Rza–Pr angle on the model using a customized goniometer Step 2: CT imaging and standard plane measurements The skull model was scanned using a CT scanner (Optima CT660, GE Healthcare, USA) with a slice thickness of 0.625 mm. The image series was aligned to the Frankfort Horizontal Plane in Horos (v3.3.6) to obtain sagittal, axial, and coronal multiplanar reconstruction (MPR) views [9]. Landmarks were identified in the respective planes using the “Point” tool. As the landmarks were located on different slices, direct angular measurement was not feasible. Therefore, slice thickness was virtually increased using the thick slab function to superimpose slices and visualize all landmarks within a single view (Fig. 2A). Angular measurements were then obtained using the “Angle” tool. These measurements represented orthographic projections and were recorded as standard sectional measurements. Step 3: reoriented plane measurements To address the limitations of sectional measurements, the MPR planes were reoriented to align with the true spatial configuration of the three landmarks defining each angle. Landmark points were marked using the “Point” tool. After increasing slice thickness with the thick slab function to visualize all landmarks simultaneously, the planes were rotated using the “Rotate” tool. The objective was to obtain an orthogonal view of the plane formed by the three landmarks. Rotation was continued in two MPR windows until the three points aligned on a single line (Supplementary Video 1). Angular values before and after reorientation are shown in Figs. 3 - A3 and B3. Fig. 3Measurement of the ∠Co–Go–Gn angle on standard radiological planes. A1: sagittal plane, A2: transverse plane, A3: coronal plane. Construction of the reoriented plane for measurement of the ∠Co–Go–Gn angle. B1: reorientation of the sagittal plane, B2: reorientation of the transverse plane, B3: final view of the completed reoriented plane and the measured angle Measurement of the ∠Co–Go–Gn angle on standard radiological planes. A1: sagittal plane, A2: transverse plane, A3: coronal plane. Construction of the reoriented plane for measurement of the ∠Co–Go–Gn angle. B1: reorientation of the sagittal plane, B2: reorientation of the transverse plane, B3: final view of the completed reoriented plane and the measured angle Data analysis Measurements obtained using the goniometer, reoriented planes, and standard anatomical planes (sagittal, coronal, and transverse) were statistically analyzed. The study included five skull models, and four angles were evaluated in each model: ∠Co–Id–Co, ∠Go–Gn–Go, ∠Co–Go–Gn, and ∠N–Rza–Pr. Each angle was measured three times by each of three independent observers. Repeated measurements for intraobserver reliability were performed at 5-day intervals, and each observer was blinded to the measurements of the others. To compare methodological approaches, both intraobserver and interobserver reliability were assessed using Intraclass Correlation Coefficients (ICC). Analyses were performed with a two-way mixed-effects model based on absolute agreement. For intraobserver reliability, ICC was calculated using the three repeated measurements of the same observer across all five skull models. For interobserver reliability, ICC was calculated using the mean of the three repeated measurements obtained by each observer. Single Measures ICC evaluated individual consistency, whereas Average Measures ICC assessed agreement among observers. Bland–Altman analyses were additionally conducted to evaluate agreement between methods and detect potential systematic differences. Results Goniometric measurements from five skulls yielded mean values of 55.8° ± 4.2° for ∠Co–Id–Co, 72.9° ± 4.9° for ∠Go–Gn–Go, 113.5° ± 6.5° for ∠Co–Go–Gn, and 36.5° ± 5.7° for ∠N–Rza–Pr. When the same angles were measured using the reoriented plane method based on CT images, similar results were observed: 55.8° ± 4.1° for the ∠Co–Id–Co angle, 72.9° ± 5.0° for the ∠Go-Gn-Go angle, 113.5° ± 6.4° for the ∠Co-Go-Gn angle, and 36.5° ± 5.7° for the ∠N-Rza-Pr angle. In sagittal sections, only ∠Co–Go–Gn (122.1° ± 8.8°) and ∠N–Rza–Pr (42.1° ± 6.3°) were measurable. In coronal sections, all angles were measurable but showed substantial deviations from reference values. In transverse sections, only ∠Co–Id–Co (68.3° ± 4.5°) and ∠Go–Gn–Go (77.5° ± 5.3°) could be identified (Table 1). Table 1Mean ± standard deviation values (in degrees) ​​of angular measurements obtained using goniometric measurement, reoriented plane, and standard sagittal, coronal, and transverse sections∠Co–Id–Co∠Go-Gn-Go∠Co-Go-Gn∠N-Rza-PrMean (°) ± SDMean (°) ± SDMean (°) ± SDMean (°) ± SDGoniometricmeasurement55.8 ± 4.272.9 ± 4.9113.5 ± 6.536.5 ± 5.7Reorientedplane55.8 ± 4.172.9 ± 5113.5 ± 6.436.5 ± 5.7Sagittalsection--122.1 ± 8.842.1 ± 6.3Coronalsection84.2 ± 9.5109.5 ± 27.4124.5 ± 7.765.2 ± 11.8Transverssection68.3 ± 4.577.5 ± 5.3-- Table 1Mean ± standard deviation values (in degrees) ​​of angular measurements obtained using goniometric measurement, reoriented plane, and standard sagittal, coronal, and transverse sections∠Co–Id–Co∠Go-Gn-Go∠Co-Go-Gn∠N-Rza-PrMean (°) ± SDMean (°) ± SDMean (°) ± SDMean (°) ± SDGoniometricmeasurement55.8 ± 4.272.9 ± 4.9113.5 ± 6.536.5 ± 5.7Reorientedplane55.8 ± 4.172.9 ± 5113.5 ± 6.436.5 ± 5.7Sagittalsection--122.1 ± 8.842.1 ± 6.3Coronalsection84.2 ± 9.5109.5 ± 27.4124.5 ± 7.765.2 ± 11.8Transverssection68.3 ± 4.577.5 ± 5.3-- Mean ± standard deviation values (in degrees) ​​of angular measurements obtained using goniometric measurement, reoriented plane, and standard sagittal, coronal, and transverse sections Goniometric measurement Reoriented plane Sagittal section Coronal section Transvers section In this study, the reliability of goniometric and radiological image–based measurements were assessed using intraclass correlation coefficients (ICC), evaluating both intraobserver consistency and interobserver agreement. Intraclass correlation coefficient (ICC) analysis results In the ICC analysis for observer reliability of goniometric measurements, the Single Measures ICC value for the first observer was calculated as 0.840 (95% CI: [0.466, 0.980]) (F(4, 8) = 21.767, p < 0.001). The second observer’s Single Measures ICC was 0.863 (95% CI: [0.527, 0.983]) (F(4, 8) = 24.367, p < 0.001). For the third observer, the Single Measures ICC value was 0.880 (95% CI: [0.580, 0.985]) (F(4, 8) = 26.236, p < 0.001). These results were statistically significant and indicate a high level of consistency among the researchers. In the ICC analysis for observer reliability of radiological measurements, the Single Measures ICC value for the first observer was calculated as 0.909 (95% CI: [0.627, 0.989]) (F(4, 8) = 25.330, p < 0.001). The second observer’s Single Measures ICC was 0.904 (95% CI: [0.609, 0.989]) (F(4, 8) = 23.869, p < 0.001). For the third observer, the Single Measures ICC value was 0.894 (95% CI: [0.576, 0.987]) (F(4, 8) = 21.425, p < 0.001). These results were statistically significant and indicate a high level of consistency among the researchers. In the interobserver agreement analysis of goniometric measurements, the Average Measures ICC was determined to be 0.922 (95% CI: [0.645, 0.991]). The result was statistically significant (F(4, 8) = 16.832, p < 0.001), indicating a high level of consistency between the observers. In the interobserver agreement analysis of the reoriented plane measurements, the Average Measures ICC was found to be 0.950 (95% CI: [0.760, 0.994]) (F(4, 8) = 17.650, p < 0.001). This result indicates that the measurements made by different researchers were highly reliable. Overall, the high ICC values obtained in all four analyses suggest that the measurements are reliable both when repeated and when performed by different individuals. Notably, the fact that the Average Measures ICC values were above 0.90 in all analyses demonstrates that reliability significantly improves when the average of multiple measurements is used (Table 2). Table 2Intraobserver and interobserver intraclass correlation coefficient (ICC) results for goniometric and CT measurementsResearcherICC ModelICC Value95% CIF (df₁ = 4, df₂ = 8) p Intraobserver ICC Analysis of Goniometric Measurements1(3,1)0.840[0.466, 0.980]21.767< 0.0012(3,1)0.863[0.527, 0.983]24.367< 0.0013(3,1)0.880[0.580, 0.985]26.236< 0.001Intraobserver ICC Analysis of Reoriented Plane Measurements1(3,1)0.909[0.627, 0.989]25.331< 0.0012(3,1)0.904[0.609, 0.989]23.869< 0.0013(3,1)0.894[0.576, 0.987]21.425< 0.001Interobserver ICC Analysis of Goniometric Measurements(3,3)0.922[0.645, 0.991]16.832< 0.001Interobserver ICC Analysis of Reoriented Plane Measurements(3,3)0.950[0.760, 0.994]17.650< 0.001 Table 2Intraobserver and interobserver intraclass correlation coefficient (ICC) results for goniometric and CT measurementsResearcherICC ModelICC Value95% CIF (df₁ = 4, df₂ = 8) p Intraobserver ICC Analysis of Goniometric Measurements1(3,1)0.840[0.466, 0.980]21.767< 0.0012(3,1)0.863[0.527, 0.983]24.367< 0.0013(3,1)0.880[0.580, 0.985]26.236< 0.001Intraobserver ICC Analysis of Reoriented Plane Measurements1(3,1)0.909[0.627, 0.989]25.331< 0.0012(3,1)0.904[0.609, 0.989]23.869< 0.0013(3,1)0.894[0.576, 0.987]21.425< 0.001Interobserver ICC Analysis of Goniometric Measurements(3,3)0.922[0.645, 0.991]16.832< 0.001Interobserver ICC Analysis of Reoriented Plane Measurements(3,3)0.950[0.760, 0.994]17.650< 0.001 Intraobserver and interobserver intraclass correlation coefficient (ICC) results for goniometric and CT measurements The results of measurements obtained at four different cross-sections were evaluated using Bland-Altman plots for four different angular measurements. Goniometer measurements were used as the reference value for the analysis. A Bland–Altman comparison between the reference and reoriented plane measurements was conducted for four craniofacial angles (Table 3). Table 3Bland–Altman analysis results for agreement between measurement methodsAnalysis parameterGoniometric measurementsXReoriented plane measurementsGoniometric measurementsXSagittal plane measurementsGoniometric measurementsXCoronal plane mesurementsGoniometric measurementsXTransverse plane∠Co–Id–CoMean bias,degree ° (95% CI)0.07(− 0.08; 0.21)-−28.33(− 31.63; −25.04)−12.27(− 13.26; −11.27)Limits of agreement,degree °−0.43; 0.57-−39.99; −16.67−15.78; −8.76Coefficient ofrepeatability(95% CI)1.01(0.92; 1.13)-23.23(20.81; 25.65)7.11(6.33; 7.89)∠Co-Go-GnMean bias,degree ° (95% CI)0.002(− 0.08; 0.09)−8.60(− 12.56; −4.91)−10.98(− 12.77; −9.19)-Limits ofagreement,degree °−0.30; 0.30−22.60; 5.41−17.31; −4.65-Coefficient ofrepeatability(95% CI)0.60(0.53; 0.68)14.00(12.71; 15.68)12.65(11.23; 14.07)-∠Go-Gn-GoMean bias,degree ° (95% CI)0.07(− 0.32; 0.46)-−36.47(− 52.18; −20.76)−4.67(− 7.13; −2.20)Limits of agreement,degree °−1.31; 1.45-−92.07; 19.14−13.39; 4.06Coefficient ofrepeatability(95% CI)2.76(2.49; 3.03)-56.7(50.84; 62.57)8.11(7.00; 9.22)∠N-Rza-PrMean bias,degree ° (95% CI)−0.023(− 0.11; 0.06)−5.56(− 6.22; −4.91)−28.65(− 31.92; −25.38)-Limits of agreement,degree °−0.33; 0.28−7.88; −3.24−40.22; −17.08-Coefficient ofrepeatability(95% CI)0.61(0.55; 0.66)4.64(4.20; 5.09)23.15(20.71; 25.60)- Table 3Bland–Altman analysis results for agreement between measurement methodsAnalysis parameterGoniometric measurementsXReoriented plane measurementsGoniometric measurementsXSagittal plane measurementsGoniometric measurementsXCoronal plane mesurementsGoniometric measurementsXTransverse plane∠Co–Id–CoMean bias,degree ° (95% CI)0.07(− 0.08; 0.21)-−28.33(− 31.63; −25.04)−12.27(− 13.26; −11.27)Limits of agreement,degree °−0.43; 0.57-−39.99; −16.67−15.78; −8.76Coefficient ofrepeatability(95% CI)1.01(0.92; 1.13)-23.23(20.81; 25.65)7.11(6.33; 7.89)∠Co-Go-GnMean bias,degree ° (95% CI)0.002(− 0.08; 0.09)−8.60(− 12.56; −4.91)−10.98(− 12.77; −9.19)-Limits ofagreement,degree °−0.30; 0.30−22.60; 5.41−17.31; −4.65-Coefficient ofrepeatability(95% CI)0.60(0.53; 0.68)14.00(12.71; 15.68)12.65(11.23; 14.07)-∠Go-Gn-GoMean bias,degree ° (95% CI)0.07(− 0.32; 0.46)-−36.47(− 52.18; −20.76)−4.67(− 7.13; −2.20)Limits of agreement,degree °−1.31; 1.45-−92.07; 19.14−13.39; 4.06Coefficient ofrepeatability(95% CI)2.76(2.49; 3.03)-56.7(50.84; 62.57)8.11(7.00; 9.22)∠N-Rza-PrMean bias,degree ° (95% CI)−0.023(− 0.11; 0.06)−5.56(− 6.22; −4.91)−28.65(− 31.92; −25.38)-Limits of agreement,degree °−0.33; 0.28−7.88; −3.24−40.22; −17.08-Coefficient ofrepeatability(95% CI)0.61(0.55; 0.66)4.64(4.20; 5.09)23.15(20.71; 25.60)- Bland–Altman analysis results for agreement between measurement methods Mean bias, degree ° (95% CI) 0.07 (− 0.08; 0.21) −28.33 (− 31.63; −25.04) −12.27 (− 13.26; −11.27) Limits of agreement, degree ° Coefficient of repeatability (95% CI) 1.01 (0.92; 1.13) 23.23 (20.81; 25.65) 7.11 (6.33; 7.89) Mean bias, degree ° (95% CI) 0.002 (− 0.08; 0.09) −8.60 (− 12.56; −4.91) −10.98 (− 12.77; −9.19) Limits of agreement, degree ° Coefficient of repeatability (95% CI) 0.60 (0.53; 0.68) 14.00 (12.71; 15.68) 12.65 (11.23; 14.07) Mean bias, degree ° (95% CI) 0.07 (− 0.32; 0.46) −36.47 (− 52.18; −20.76) −4.67 (− 7.13; −2.20) Limits of agreement, degree ° Coefficient of repeatability (95% CI) 2.76 (2.49; 3.03) 56.7 (50.84; 62.57) 8.11 (7.00; 9.22) Mean bias, degree ° (95% CI) −0.023 (− 0.11; 0.06) −5.56 (− 6.22; −4.91) −28.65 (− 31.92; −25.38) Limits of agreement, degree ° Coefficient of repeatability (95% CI) 0.61 (0.55; 0.66) 4.64 (4.20; 5.09) 23.15 (20.71; 25.60) For the ∠Co–Id–Co angle, the analysis revealed minimal bias (0.07°, 95% CI: −0.08 to 0.21), with narrow limits of agreement ranging from − 0.44° (95% CI: −0.69 to − 0.19) to 0.57° (95% CI: 0.32 to 0.82). Repeatability was high, with a coefficient of repeatability of 1.01 (95% CI: 0.92 to 1.13). For the ∠Co–Go–Gn angle, a mean bias of 0.002° (95% CI: −0.08 to 0.09) was observed, with narrow limits of agreement from − 0.30° (95% CI: −0.45 to − 0.15) to 0.30° (95% CI: 0.15 to 0.45), and a coefficient of repeatability of 0.60° (95% CI: 0.53 to 0.68). In the case of the ∠Go–Gn–Go angle, the mean bias was 0.07° (95% CI: −0.32 to 0.46), with acceptable limits of agreement extending from − 1.31° (95% CI: −1.64 to − 0.99) to 1.45° (95% CI: 1.12 to 1.77), and a coefficient of repeatability of 2.76° (95% CI: 2.49 to 3.03). Finally, for the ∠N–Rza–Pr angle, the mean bias was − 0.023° (95% CI: −0.11 to 0.06), with narrow limits of agreement from − 0.33° (95% CI: −0.40 to − 0.26) to 0.28° (95% CI: 0.21 to 0.35), and a coefficient of repeatability of 0.61° (95% CI: 0.55 to 0.66) (Fig. 4). Fig. 4Bland–Altman plots showing the agreement between goniometric measurements and reoriented plane measurements for four angles (∠Co–Id–Co, ∠Go–Gn–Go, Co–Go–Gn, ∠N–Rza–Pr) Bland–Altman plots showing the agreement between goniometric measurements and reoriented plane measurements for four angles (∠Co–Id–Co, ∠Go–Gn–Go, Co–Go–Gn, ∠N–Rza–Pr) In sagittal sections, the ∠Co–Id–Co and ∠Go–Gn–Go angles could not be identified; therefore, these angles were not measurable in sagittal images. For the ∠Co–Go–Gn angle, Bland–Altman analysis comparing the sagittal plane measurement with the reference value showed a mean bias of − 8.60° (95% CI: −12.56 to − 4.91). The limits of agreement were wide, ranging from − 22.60° (95% CI: −29.52 to − 15.68) to 5.41° (95% CI: −1.51 to 12.33), and the coefficient of repeatability was 14.00° (95% CI: 12.71 to 15.68). For the ∠N–Rza–Pr angle, Bland–Altman analysis demonstrated a mean bias of − 5.56° (95% CI: −6.22 to − 4.91), with narrow limits of agreement from − 7.88° (95% CI: −9.03 to − 6.74) to − 3.24° (95% CI: −4.39 to − 2.09). The coefficient of repeatability was 4.64° (95% CI: 4.20 to 5.09) (Fig. 5). Fig. 5Bland–Altman plots showing the agreement between goniometric measurements and sagittal sections measurements for two angles (∠Go–Gn–Go, ∠N–Rza–Pr) Bland–Altman plots showing the agreement between goniometric measurements and sagittal sections measurements for two angles (∠Go–Gn–Go, ∠N–Rza–Pr) In coronal sections, all four angular measurements could be identified. Bland–Altman analysis for the ∠Co–Id–Co angle showed a mean bias of − 28.33° (95% CI: −31.63 to − 25.04), with wide limits of agreement ranging from − 39.99° (95% CI: −45.75 to − 34.23) to − 16.67° (95% CI: −22.44 to − 10.91). The coefficient of repeatability was 23.23° (95% CI: 20.81 to 25.65). For the ∠Co–Go–Gn angle, the mean bias was − 10.98° (95% CI: −12.77 to − 9.19), with moderately wide limits of agreement from − 17.31° (95% CI: −20.43 to − 14.18) to − 4.65° (95% CI: −7.78 to − 1.53). The coefficient of repeatability was 12.65° (95% CI: 11.23 to 14.07). Regarding the ∠Go–Gn–Go angle, a mean bias of − 36.47° (95% CI: −52.18 to − 20.76) was observed, with extremely wide limits of agreement ranging from − 92.07° (95% CI: −119.54 to − 64.60) to 19.14° (95% CI: −8.34 to 46.61), and a coefficient of repeatability of 56.71° (95% CI: 50.84 to 62.57). Finally, for the ∠N–Rza–Pr angle, the analysis showed a mean bias of − 28.65° (95% CI: −31.92 to − 25.38), with wide limits of agreement ranging from − 40.22° (95% CI: −45.93 to − 34.50) to − 17.08° (95% CI: −22.79 to − 11.36). The coefficient of repeatability was 23.15° (95% CI: 20.71 to 25.60) (Fig. 6). Fig. 6Bland–Altman plots showing the agreement between goniometric measurements and coronal sections measurements for four angles (∠Co–Id–Co, ∠Go–Gn–Go, ∠N–Rza–Pr, Co–G0–Gn) Bland–Altman plots showing the agreement between goniometric measurements and coronal sections measurements for four angles (∠Co–Id–Co, ∠Go–Gn–Go, ∠N–Rza–Pr, Co–G0–Gn) In transverse sections, the ∠Co–Go–Gn and ∠N–Rza–Pr angles could not be identified; therefore, these angles were not measurable in transverse images. For the ∠Co–Id–Co angle, Bland–Altman analysis showed a mean bias of − 12.27° (95% CI: −13.26 to − 11.27), with moderate limits of agreement ranging from − 15.78° (95% CI: −17.70 to − 14.04) to − 8.76° (95% CI: −10.49 to − 7.02). The coefficient of repeatability was 7.11° (95% CI: 6.33 to 7.89). For the ∠Go–Gn–Go angle, Bland–Altman analysis revealed a mean bias of − 4.67° (95% CI: −7.13 to − 2.20), with moderately wide limits of agreement from − 13.39° (95% CI: −17.70 to − 9.08) to 4.06° (95% CI: −0.25 to 8.37). The coefficient of repeatability was 8.11° (95% CI: 7.00 to 9.22) (Fig. 7). Fig. 7Bland–Altman plots showing the agreement between goniometric measurements and sagittal sections measurements for two angles (∠Go–Gn–Go, ∠N–Rza–Pr) Bland–Altman plots showing the agreement between goniometric measurements and sagittal sections measurements for two angles (∠Go–Gn–Go, ∠N–Rza–Pr) Discussion This study aimed to improve the accuracy of angular measurements in three-dimensional radiological images by addressing the limitations of conventional section-based methods. Measurements obtained using the reoriented plane approach closely matched direct goniometric reference values, whereas standard sagittal, coronal, and transverse sections showed substantial deviations, reflecting the limitations of single-plane assessments in representing three-dimensional geometry. By aligning the measurement plane with the three landmarks defining each angle, the reoriented method enabled accurate geometric visualization, resulting in minimal bias and narrow limits of agreement in Bland–Altman analyses. Intraclass correlation coefficients also demonstrated high intraobserver and interobserver reliability. Together, these findings emphasize that plane alignment with the true spatial configuration of landmarks is essential for precise and reproducible angular measurement in three-dimensional datasets. Angular measurements derived from standard sagittal, coronal, and transverse sections are inherently limited in three-dimensional anatomy, especially when landmarks lie in different spatial planes [10, 11]. In these situations, single-plane measurements represent orthographic projections rather than true three-dimensional angles, resulting in systematic error [10–12]. This limitation was evident in the present study, as some angles could not be reliably identified or measured in certain sectional views. Projection-related errors are a major source of inaccuracy, as out-of-plane landmark displacement can lead to over- or underestimation of angular values. These errors are particularly evident in anatomically complex regions such as the craniofacial skeleton, where curved structures and asymmetric landmark distribution are common [6, 13–17]. Moreover, the inability to visualize all defining landmarks within a single section limits the applicability of sectional methods and increases variability across studies using different plane orientations. These limitations have important clinical and research implications. Inaccurate angular measurements may compromise surgical planning, morphometric analyses, and the design of patient-specific reconstruction implants that depend on precise three-dimensional geometry [14, 18–20]. Plane-dependent variability also reduces comparability between studies and hinders methodological standardization. These issues highlight the need for measurement strategies that move beyond conventional sectional approaches and accurately reflect true three-dimensional anatomy. The reoriented plane approach addresses the geometric limitations of sectional measurements by aligning the measurement plane with the true spatial configuration of the three defining landmarks. By visualizing all landmarks within a single plane, it enables direct measurement of the actual three-dimensional angle rather than an orthographic projection, thereby improving accuracy. In this study, reoriented plane measurements demonstrated minimal bias and narrow limits of agreement relative to direct goniometric values, indicating high validity [21]. Unlike sectional measurements, which are sensitive to plane orientation and prone to variability, the reoriented plane method produced consistent results across repeated assessments and observers. Its validity is further supported by its geometric basis [10]. In three-dimensional space, an angle is defined by three non-collinear points and lies within the plane they form; measurements outside this plane are inherently affected by projection distortion [10, 22]. By aligning the reconstruction plane with the landmark-defined plane, the reoriented method preserves true angular geometry and eliminates out-of-plane error. These findings indicate that plane reorientation is essential for accurate angular assessment in three-dimensional radiological datasets, particularly in anatomically complex regions. Reliable angular measurement is critical for the effective use of three-dimensional radiological data in clinical and research settings. In this study, reliability was evaluated using intraclass correlation coefficients (ICC) for both intraobserver and interobserver agreement. Although all methods demonstrated high reliability, the highest ICC values were consistently observed with the reoriented plane approach. By enabling precise landmark identification, the reoriented method improved methodological consistency and reproducibility. Average Measures ICC values above 0.90 further indicate enhanced reliability when multiple measurements are combined. These findings support the suitability of the reoriented plane approach for longitudinal and multicenter studies requiring consistent angular assessment. Previous studies assessing angular measurements in three-dimensional radiological images have primarily relied on standard anatomical planes, including sagittal, coronal, and transverse Sects [23–26]. Although practical, these single-plane approaches may inadequately represent complex three-dimensional angular relationships, particularly when landmarks lie in different spatial orientations [22]. Thus, inconsistencies in reported angular values likely result from differences in plane selection. Most prior efforts to improve measurement accuracy have focused on landmark definition, imaging resolution, or software tools, while the geometric validity of the measurement plane itself has received limited attention [23–26]. Consequently, discrepancies between studies and a lack of methodological standardization persist . In contrast, this study systematically compared standard sectional measurements with a geometrically aligned reoriented plane, using direct goniometric values as the reference. Reoriented measurements closely matched goniometric results, whereas sectional measurements showed substantial deviations, providing a methodological explanation for inconsistencies in the literature. These findings emphasize that plane alignment with the true spatial configuration of landmarks is essential for accurate and reproducible angular assessment in three-dimensional datasets. A major strength of this study is the development of a custom three-dimensional goniometric ruler, which enabled accurate angular measurement on complex craniofacial structures not reliably assessable with conventional two-dimensional devices. This instrument served as a robust physical reference standard for validating radiological measurements. Another strength is the use of anatomically accurate plastic skull models, which minimized variability from deformation or erosion and ensured consistent landmark identification. The inclusion of multiple observers and repeated measurements enabled rigorous evaluation of intraobserver and interobserver reliability, supported by complementary ICC and Bland–Altman analyses. Finally, the systematic comparison of goniometric, sectional, and reoriented plane measurements within a single framework enhanced internal validity and strengthened the robustness of the conclusions. Accurate angular measurements from three-dimensional radiological images are essential for clinical and research applications, particularly in craniofacial analysis and reconstructive surgery [5, 8, 18, 27]. Inaccurate angular assessment may result in geometric errors in surgical planning, implant design, and postoperative evaluation. Our findings show that conventional section-based methods can introduce clinically relevant inaccuracies, whereas the reoriented plane approach more faithfully represents true anatomical geometry. In clinical practice, precise angular data are crucial for the design and placement of patient-specific craniofacial implants [14, 20, 28]. Errors in angular measurement may impair implant fit, disrupt biomechanical load distribution, and increase postoperative risk. By enhancing accuracy and reproducibility, the reoriented plane approach can improve preoperative planning and implant congruence. From a research perspective, geometrically valid measurement planes enhance standardization, improve comparability across studies, and support longitudinal and multicenter analyses. The proposed framework is also compatible with automated and AI-based workflows, reducing operator dependency and improving measurement robustness [29, 30]. Limitations and future directions This study has several limitations. Although anatomically accurate plastic skull models ensured methodological standardization, they do not fully represent in vivo conditions, as the absence of soft tissues and physiological variability may affect landmark identification. Therefore, extrapolation to clinical datasets should be made cautiously. Although the reoriented plane approach showed high accuracy and reliability under controlled conditions, its performance across different imaging modalities, software platforms, and anatomical regions requires further validation. Future research should evaluate clinical datasets, pathological cases, and integration into automated or semi-automated systems. Conclusions This study demonstrates that accurate angular assessment in three-dimensional radiological imaging requires alignment of the measurement plane with the true spatial configuration of anatomical landmarks. Conventional section-based measurements confined to standard anatomical planes may introduce substantial inaccuracies due to projection-related distortion and incomplete landmark visualization. In contrast, the reoriented plane approach provides a geometrically valid and reproducible method that closely corresponds to direct goniometric reference values and demonstrates high intraobserver and interobserver reliability. Overall, this method offers a robust framework for improving the accuracy, reproducibility, and standardization of angular measurements, with potential benefits for morphometric analyses and patient-specific craniofacial surgical planning. Supplementary Information Supplementary Material 1. Supplementary Material 1.