Age Estimation with the Greulich-Pyle Atlas using 3T MR Images of Hand and Wrist
Thomas Widek1,2, Pia Genet3,4, Thomas Ehammer1, Eva Scheurer5, and Thorsten Schwark1,6

1Ludwig Boltzmann Institute for Clinical Forensic Imaging, Graz, Austria, 2Medical University of Graz, Graz, Austria, 3University Centre of Legal Medicine, Lausanne University Hospital, Lausanne, Switzerland, 4University Centre of Legal Medicine, Geneva University Hospital, Geneva, Switzerland, 5Institute of Forensic Medicine, University of Basel - Health Department Basel-Stadt, Basel, Switzerland, 6Department of Forensic Medicine, Laboratoire national de santé, Dudelange, Luxembourg


Bone age estimation of the hand is very common both in a forensic context and for clinical purposes. Currently, this is done by assessing plain X-rays of the hand. This is a controversial issue, especially in the forensic context, as legal proceedings lack a medical indication for the use of ionizing radiation. The aim of the current study was to validate the use of the X-ray based Greulich-Pyle atlas method on hand MR images in a healthy male cohort. The results show that the application of the Greulich-Pyle method is feasible and that it can be used in daily routine.


Bone age determination of the hand by means of X-ray examination represents an important pillar of forensic age estimation. However, the associated radiation exposure is still subject to a controversial discussion. Therefore, the search for radiation-free alternatives such as MRI is in the focus of forensic research. Recently, two pilot studies1,2 have investigated the applicability of the X-ray based atlas method of Greulich and Pyle3 (GP) with MR hand images. The aim of the current study was the validation of the use of the GP-method with 3T MR images in a large male cohort.

Materials and Methods

228 healthy male subjects (age 13.01 – 20.99 years, mean 17.29 ± 2.28y) underwent an MRI of the left hand and wrist. The investigations were performed with a 3T MR (Trio, a TIM system & Skyra, Siemens Healthcare) and the volunteers were placed in prone position with the arm outstretched; the hand was fixed and the standard head and neck coil was used. The following sequences in coronal orientation were applied: 3D T1w VIBE (TR/TE 14/4.01ms, 0.9x0.9x0.9mm³) and 3D T2w DESS (TR/TE 14.28/5.18ms, 0.8.x0.8x0.8mm³). The MR images were evaluated independently by two blinded readers (one with experience and one inexperienced) according to the defined standards of the GP atlas3. The age estimates of both readers were compared to each other and to chronological age using Bland-Altman4 and regression plots. The assessed age was also compared to the standard deviations as provided in the GP atlas3 (not available SDs were assumed). Intra-observer and inter-observer agreements were calculated using weighted Cohen’s Kappa5.


Fig.1 shows an example of MR images of the hand and wrist of a 15 years old male.The Bland-Altman plot in Fig.2 displays the comparison between the two readers. A good agreement with almost zero mean difference (dmean = 0.15y) was observed. The limits of agreement were about ± 1 year (LOA: 1.27y, -0.98y). The weighted Cohen’s Kappa for the inter-observer and the intra-observer agreement were very good with κw=0.85 and κw=0.88, respectively. Fig.3a and b show Bland-Altman-like plots of the difference between chronological age and assessed age for both readers. Both plots show no systematic deviation and a small mean underestimation (d1mean = 0.40y, d2mean = 0.54y). The red and blue lines represent the single and double standard deviations as shown in the Greulich-Pyle atlas3 for each age group. Both readers assessed about 50% of the data correctly. Sixty-five percent of the assessments were within one standard deviation and about 97% within two standard deviations. The regression plots (Fig.4 a, b) show good linear correlations for the assessed age with the chronological age (y1 = 0.94x + 0.81, R = 0.86; y2 = 0.95x + 0.78, R = 0.88).

Discussion and Conclusion

Age estimation of hand MRIs using the GP method shows a very good inter-observer agreement. This is especially notable as one reader was inexperienced. The high agreements shows that the method is reliable. The variation concerning the results of the age assessments is similar to the study shown in the Greulich-Pyle atlas3. This makes the assessment of the MR images comparable to the assessment of X-ray images. Thus, the use of MRI of the hand and wrist as radiation-free method for the estimation of skeletal age is ready for daily forensic practice. However, in order to reduce the variation, multifactorial assessment based on examinations of multiple body regions, i.e. hand and wrist, wisdom teeth and clavicles, is still recommended.


No acknowledgement found.


1. Hojreh A, Gamper J, Schmook MT, et al. Hand MRI and the Greulich-Pyle atlas in skeletal age estimation in adolescents. Skeletal Radiol. 2018;47(7):963-971.

2. Urschler M, Krauskopf A, Widek T, et al. Applicability of Greulich-Pyle and Tanner-Whitehouse grading methods to MRI when assessing hand bone age in forensic age estimation: A pilot study. Forensic Sci Int. 2016;266:281-288.

3. Greulich WW, Pyle SI. Radiographic Atlas of Skeletal Development of Hand and Wrist. 2nd ed. Stanford: Stanford University Press; 1959.

4. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1(8476):307-310.

5. J. Cohen, Weighted kappa: nominal scale agreement with provision for scaled disagreement or partial credit. Psychol Bull. 1968; 70 (4) 213-20.


Fig.1: MR images of the left hand and wrist of a 15 years old boy (GP standard 26). [3D T1w VIBE (left side). 3D T2w DESS (right side)].

Fig.2: Bland-Altman plot of the two readers. The middle bold dashed line is the mean difference and the two outer bold dashed lines define the limits of agreement. Dotted lines represent lower and upper limits of 95 % CI of bold dashed lines

Fig.3 a+b: Plots of the difference of the estimated age to the chronological age. The middle bold line is the mean difference and the two outer dashed lines define the limits of agreement. The red and blue lines represent the single and double standard deviations (as reported by Greulich and Pyle3)

Fig.4 a+b: Regression plots of both readers. The grey dotted line indicates the identity line, the bold line is the regression line, and the dot-dashed red line represents the eighteen years border

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)