Research Paper
Choice of head circumference formula shifts ultrasound gestational age estimates by more than a week in late pregnancy
1Department of Clinical Disciplines, Jalal-Abad International University, 720906, 24 Kurmanbek Str., Manas, Kyrgyzstan
2Department of Morphological Disciplines, Jalal-Abad International University, 720906, 24 Kurmanbek Str., Manas, Kyrgyzstan
* Corresponding author: Zhumanalieva Maksudakan, maksudabekievna68@gmail.com
Abstract
Background: Fetal head circumference (HC) is a core obstetric ultrasound measurement for dating and growth monitoring, yet systems and charts calculate it differently and head shape varies.
Objective: To quantify the effect of the perimeter formula on HC and dating, describe head shape across pregnancy, and examine image acquisition practice.
Methods: We analysed 999 two-dimensional fetal head ultrasound images from the openly licensed HC18 dataset, each with an expert-annotated skull ellipse, reference HC and pixel spacing. The minor and major axes of fitted ellipses were taken as biparietal (BPD) and occipitofrontal (OFD) diameters. HC was recalculated using the Ramanujan perimeter, π/2 × (BPD + OFD) and 1.62 × (BPD + OFD). Gestational age was estimated with the Hadlock HC equation.
Results: HC ranged from 44 to 346 mm (11.4–40.1 weeks); 70.2% of images were from the second trimester, concentrated at 18–22 weeks (53.5%). Ramanujan HC agreed closely with the reference (bias +0.8 mm), and the π/2 approximation was within 0.6 mm of it. The 1.62 approximation overestimated HC by 4.8 mm (2.8%), shifting gestational age by +0.9, +2.8 and +7.8 days in the first, second and third trimesters (maximum 11.6 days). Median cephalic index was 79.5; 15.8% of heads were dolichocephalic (<75) and 11.3% brachycephalic (>85), varying by trimester (p < 0.001). Pixel size rose with HC (ρ = 0.81), indicating that sonographers adjust depth and zoom as the head grows.
Conclusions: The 1.62 approximation can overdate late pregnancies by about a week. Head shape should be considered when BPD alone is used for dating. Reporting the formula used is essential for comparing measurements across systems and charts.
Keywords: Fetal BiometryHead CircumferenceBiparietal DiameterCephalic IndexGestational AgeHadlock Formula
1. Introduction
Ultrasound biometry of the fetal head is one of the most frequently performed measurements in medicine. Head circumference (HC) and biparietal diameter (BPD) are used to date pregnancies when the last menstrual period is uncertain or when the first-trimester crown–rump length is unavailable, to monitor growth and to detect abnormalities such as microcephaly [1], [2]. International guidelines define how the transthalamic plane should be obtained and how the calipers should be placed [1], [3], and reference charts for HC have been published from several populations, including the multinational INTERGROWTH-21st standards [4], [5], [6].
Less attention has been paid to how HC is calculated once the head has been outlined. The skull in the transthalamic plane is approximately an ellipse, whose perimeter has no simple exact formula. Ultrasound systems and published charts have used either an ellipse fit with a numerical or series approximation of the perimeter, or simple formulae based on the sum of the two diameters, typically π/2 × (BPD + OFD) or 1.62 × (BPD + OFD) [5]. These approaches give different results, particularly for large heads, and the differences propagate into gestational-age estimates. Head shape adds further variation: dolichocephalic and brachycephalic heads alter the relationship between BPD and gestational age, which is why HC is generally preferred to BPD alone for dating [1], [7].
Many of these questions can be answered with real images. The HC18 dataset, released under an open licence by Radboud University Medical Center, contains ultrasound images of the fetal head in the standard plane across all trimesters, with expert annotations and pixel calibration [8], [9]. This study used it to (i) quantify the agreement between perimeter formulae and their effect on Hadlock gestational-age estimates, (ii) describe fetal head shape across pregnancy using the cephalic index, and (iii) describe acquisition characteristics, such as magnification and the size of the head on the screen, that affect measurement precision.
2. Methodology
2.1 Data Source
The training set of the HC18 dataset was analysed [8], [9]. It contains 999 two-dimensional ultrasound images of the fetal head in the standard plane, acquired during routine examinations at Radboud University Medical Center (Nijmegen, the Netherlands) by experienced sonographers, with an annotated ellipse outlining the skull and a reference HC in millimetres, together with the pixel size of each image. The dataset is distributed under the Creative Commons Attribution 4.0 licence. The images are de-identified, no clinical information is provided and no new data were collected, so ethical approval was not required. Figure 1 shows real images from each trimester.
2.2 Ellipse Measurements
For each image, the outer contour of the filled annotation mask was extracted and an ellipse was fitted by least squares with OpenCV. The minor and major axes of the fitted ellipse were converted to millimetres with the pixel size and taken as BPD and OFD analogues; because they are measured on the annotated skull ellipse, they correspond to diameters measured along the ellipse outline rather than to the outer-to-inner caliper convention used clinically for BPD. HC was then calculated in three ways: (1) with Ramanujan's approximation of the ellipse perimeter, π[3(a + b) − √((3a + b)(a + 3b))], where a and b are the semi-axes [10], which is accurate to a small fraction of a millimetre for fetal head shapes; (2) as π/2 × (BPD + OFD); and (3) as 1.62 × (BPD + OFD). The cephalic index (CI) was calculated as BPD/OFD × 100 and classified as dolichocephalic (< 75), mesocephalic (75–85) or brachycephalic (> 85) [1], [7]. The fraction of the image occupied by the head and whether the annotation touched the image border were also recorded.
2.3 Gestational Age
Gestational age (GA) was estimated from the reference HC with the Hadlock equation [11]: GA (weeks) = 8.96 + 0.540 × HC + 0.0003 × HC³, with HC in centimetres. Images were assigned to the first (< 14 weeks), second (14 to < 28 weeks) or third trimester (≥ 28 weeks). The effect of the HC formula on dating was calculated as the difference between the GA derived from each approximation and that derived from the Ramanujan HC, expressed in days.
2.4
Statistical analysis
Agreement between HC methods was assessed with Bland–Altman analysis (bias and 95% limits of agreement) [12]. Associations were assessed with Spearman correlation and differences between trimesters with the Kruskal–Wallis test. Analyses were performed in Python with OpenCV, NumPy, pandas and SciPy.
3. Results and Disccussion
3.1 Study Images
The reference HC ranged from 44.3 to 346.4 mm (median 174.1 mm), corresponding to a Hadlock GA of 11.4 to 40.1 weeks (median 19.9). There were 154 first-trimester (15.4%), 701 second-trimester (70.2%) and 144 third-trimester images (14.4%) (Figure 2a). More than half of all images (53.5%) fell between 18 and 22 weeks, consistent with the routine mid-trimester anomaly scan [3]. Most images measured 800 × 540 pixels, and the head ellipse touched the image border in only five. Table 1 summarises the biometry by GA band; BPD and OFD increased in parallel with HC (Figure 2b), and BPD was almost perfectly linearly related to HC (r = 0.996; BPD ≈ 0.280 × HC + 0.2 mm).
| GA Band, Weeks | Images | HC, Mm | Short Axis (BPD), Mm | Long Axis (OFD), Mm | Cephalic Index | Pixel Size, Mm |
|---|---|---|---|---|---|---|
| 10–14 | 154 | 75.3 | 21.7 | 26.3 | 82.8 | 0.077 |
| 14–18 | 90 | 105.0 | 30.2 | 37.4 | 81.9 | 0.100 |
| 18–22 | 534 | 174.5 | 48.9 | 62.2 | 78.4 | 0.128 |
| 22–26 | 59 | 219.1 | 61.1 | 77.8 | 78.8 | 0.163 |
| 26–30 | 50 | 260.7 | 73.6 | 92.5 | 79.5 | 0.187 |
| 30–34 | 70 | 288.5 | 81.7 | 102.0 | 79.8 | 0.219 |
| 34–42 | 42 | 316.1 | 90.0 | 114.2 | 80.4 | 0.232 |
3.2 Agreement between HC Formulae
The Ramanujan perimeter of the fitted ellipse agreed closely with the reference HC (bias +0.78 mm; 95% limits of agreement +0.20 to +1.37 mm), and the π/2 × (BPD + OFD) approximation was even closer (bias +0.15 mm; −0.58 to +0.89 mm) (Table 2). In contrast, the 1.62 × (BPD + OFD) formula overestimated HC by 5.6 mm on average relative to the reference (limits +1.5 to +9.7 mm), or 3.3%. Compared with the Ramanujan perimeter, the bias of the 1.62 formula was +4.8 mm (2.8%) and increased proportionally with head size, from +2.2 mm in the first to +8.2 mm in the third trimester (Figure 3a, Table 3).
| Formula | Bias, Mm | 95% Limits of Agreement, Mm | Mean % Bias |
|---|---|---|---|
| Ramanujan perimeter of fitted ellipse | +0.78 | +0.20 to +1.37 | +0.47 |
| π/2 × (BPD + OFD) = 1.57 × (BPD + OFD) | +0.15 | -0.58 to +0.89 | +0.12 |
| 1.62 × (BPD + OFD) | +5.62 | +1.54 to +9.70 | +3.25 |
3.3 Effect on Gestational Dating
These differences translated into systematic dating errors (Table 3, Figure 3b). Using the 1.62 formula instead of the ellipse perimeter increased the estimated GA by 0.9 days in the first trimester, 2.8 days in the second and 7.8 days in the third, with a maximum of 11.6 days. The π/2 formula reduced the estimated GA by less than one day on average in every trimester (maximum 3.6 days).
| Trimester | Images | Median GA, Wk | Median HC, Mm | 1.62 Formula − Ramanujan, Mm | Dating Shift, Days (1.62) | Dating Shift, Days (1.57) | Cephalic Index | Head Area, % of Image |
|---|---|---|---|---|---|---|---|---|
| First (<14 wk) | 154 | 13.2 | 75 | +2.2 | +0.9 | -0.1 | 82.8 | 16 |
| Second (14–28 wk) | 701 | 19.9 | 174 | +4.7 | +2.8 | -0.4 | 78.9 | 33 |
| Third (≥28 wk) | 144 | 32.0 | 290 | +8.2 | +7.8 | -0.9 | 79.7 | 33 |
3.4 Head Shape
The cephalic index had a median of 79.5 (interquartile range 76.5–82.4; mean 79.5 ± 4.6). Overall, 72.9% of heads were mesocephalic, 15.8% dolichocephalic and 11.3% brachycephalic (Table 4, Figure 4). Head shape varied with gestation (Kruskal–Wallis p < 0.001; ρ with GA = −0.19): first-trimester heads were rounder (median CI 82.8, 27.9% brachycephalic), second-trimester heads were the most elongated (median 78.9, 19.0% dolichocephalic), and third-trimester heads were intermediate (79.7). Figure 5 shows real examples of the three shape categories.
| Trimester | Images | Dolichocephalic (CI < 75), % | Mesocephalic (75–85), % | Brachycephalic (CI > 85), % |
|---|---|---|---|---|
| First (<14 wk) | 154 | 5.2 | 66.9 | 27.9 |
| Second (14–28 wk) | 701 | 19.0 | 73.0 | 8.0 |
| Third (≥28 wk) | 144 | 11.8 | 78.5 | 9.7 |
| All | 999 | 15.8 | 72.9 | 11.3 |
3.5 Image Acquisition
The pixel size ranged from 0.049 to 0.393 mm and increased strongly with HC (ρ = 0.81, p < 0.001), from a median of 0.077 mm in the first to 0.221 mm in the third trimester (Figure 6a). The head occupied a median of 31% of the image (5th–95th percentile 10–49%), less in the first trimester (16%) than later (33%) (Figure 6b). Thus, sonographers widened the field of view as the head grew, which kept the head at a similar size on screen after the first trimester but meant that each pixel represented a larger distance late in pregnancy.
3.6 Sensitivity of Dating to Measurement Error
Because the Hadlock equation is non-linear, the same absolute HC error has a larger effect on dating later in pregnancy (Table 5, Figure 7). A 1 mm error in HC changed the estimated GA by 0.42 days at 13 weeks, 0.63 days at 22 weeks and 1.1 days at 39 weeks. Combined with the larger pixel size used late in pregnancy, a one-pixel caliper error in both diameters changed the estimate by about 0.1 day in the first trimester but 0.65–0.86 days near term. For a head of typical shape (CI 80), the 1.62 formula shifted the estimate by +1 day at an HC of 80 mm, +3.6 days at 200 mm and +10.8 days at 340 mm, whereas the π/2 formula shifted it by at most −1.2 days.
| HC, Mm | GA, Weeks | BPD / OFD, Mm | Shift with 1.62 Formula, Days | Shift with π/2 Formula, Days | Dating Error per 1 Mm HC, Days | Typical Pixel Size, Mm | Dating Error per 1-pixel Caliper Error, Days |
|---|---|---|---|---|---|---|---|
| 80 | 13.4 | 22.6 / 28.2 | +0.9 | -0.1 | 0.42 | 0.079 | 0.10 |
| 100 | 14.7 | 28.2 / 35.3 | +1.2 | -0.1 | 0.44 | 0.093 | 0.13 |
| 150 | 18.1 | 42.3 / 52.9 | +2.2 | -0.2 | 0.52 | 0.113 | 0.18 |
| 200 | 22.2 | 56.4 / 70.5 | +3.6 | -0.4 | 0.63 | 0.157 | 0.31 |
| 250 | 27.1 | 70.5 / 88.1 | +5.5 | -0.6 | 0.77 | 0.178 | 0.43 |
| 300 | 33.3 | 84.6 / 105.8 | +8.1 | -0.9 | 0.94 | 0.219 | 0.65 |
| 340 | 39.1 | 95.9 / 119.9 | +10.8 | -1.2 | 1.11 | 0.246 | 0.86 |
3.7 Real Images across Gestation
Figure 8 shows annotated real images at approximately four-week intervals from 12 to 36 weeks. Across gestation, the skull outline and intracranial anatomy become progressively better defined, and the field of view widens, as reflected in the increasing pixel size.
3.8 Discussion
This analysis of 999 real fetal head ultrasound images shows that the method used to compute HC from the skull outline has a measurable effect on dating. The ellipse perimeter and the π/2 × (BPD + OFD) approximation agreed within about a millimetre, but the 1.62 × (BPD + OFD) formula overestimated HC by about 3%, which shifted the Hadlock GA by a mean of 7.8 days, and by up to 11.6 days, in the third trimester.
The reason is geometric. For an ellipse, π/2 × (BPD + OFD) is the perimeter of a circle with the mean of the two diameters and slightly underestimates the true perimeter, whereas a factor of 1.62 overestimates it for any realistic fetal head shape. Because the error is proportional, it is small in absolute terms early in pregnancy, when dating is most accurate and most often performed, but grows with head size. Although HC-based dating is not recommended in the third trimester [1], [2], HC is routinely compared with growth charts at that stage, and a systematic 3% overestimate would shift centiles. Measurements made with different formulae should therefore not be compared directly, and growth charts should be applied with the formula used in their construction [4], [5]. Reports and research articles should state how HC was calculated.
The cephalic index distribution was consistent with established knowledge that the normal fetal head is moderately elongated, with roughly three-quarters of heads in the 75–85 range [1], [7]. The rounder shape in the first trimester and the more elongated shape in mid-pregnancy agree with reports that the CI changes with gestation [13]. About one in four heads fell outside the conventional range. In such heads, BPD alone underestimates (dolichocephaly) or overestimates (brachycephaly) gestational age, whereas HC is relatively independent of shape; this is the main reason for preferring HC to BPD in dating and growth assessment [1], [7]. Breech presentation, oligohydramnios and fetal position can produce dolichocephaly without pathological significance, but persistent or extreme values may warrant a careful look at the skull and brain.
The concentration of images at 18–22 weeks reflects routine practice in many countries, where the mid-trimester anomaly scan is the main scan of pregnancy [3], and suggests that the dataset mirrors real clinical workload rather than a designed uniform sample. The strong increase in pixel size with HC shows that sonographers adjust depth and zoom to keep the head at a similar size on screen. The precision of a caliper placement is limited by pixel size, so a one-pixel error corresponds to about 0.08 mm in the first trimester but 0.22 mm in the third; in practice, observer variability, which is of the order of several millimetres for HC, is far larger [14]. Because the Hadlock equation is steeper late in pregnancy, each millimetre of error costs about two and a half times as many days at term as in the first trimester (1.1 vs 0.4 days; Table 5), which is one reason why dating should be based on the earliest reliable measurement [1], [2].
3.9 Strengths and caveats
The strengths of this study include the use of real clinical ultrasound images with expert annotations and pixel calibration, a large sample across all trimesters, and transparent calculations that can be reproduced from openly licensed data.
Several limitations should be noted. The dataset contains no information on true gestational age, maternal characteristics or outcomes, so GA was estimated from HC itself and the dating differences are relative, not absolute, errors. The BPD and OFD analogues were derived from the annotated ellipse and do not follow the clinical outer-to-inner caliper convention for BPD, so their absolute values differ slightly from clinical measurements, although the CI and the formula comparisons are unaffected. The annotations were made by a small number of experts at one centre, and inter-observer variability could not be assessed. Images from the same pregnancy may appear more than once, and the number of women is not reported in the image-level data. Finally, images were acquired with machines from one manufacturer, so acquisition characteristics may differ elsewhere.
4. Conclusion
In 999 real fetal head ultrasound images, the ellipse perimeter and the π/2 × (BPD + OFD) approximation gave nearly identical HC values, whereas the 1.62 × (BPD + OFD) formula overestimated HC by about 3% and shifted gestational-age estimates by up to 11.6 days in late pregnancy. About one in four fetal heads fell outside the conventional mesocephalic range, with rounder heads early and more elongated heads in mid-pregnancy. The HC formula should be reported and matched to the reference chart, and HC rather than BPD alone should be preferred when head shape is atypical.
Acknowledgment
The authors thank Radboud University Medical Center and the creators of the HC18 dataset for making the annotated fetal ultrasound images openly available.
Funding Information
This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.
Author Contributions Statement
| Name of Author | C | M | So | Va | Fo | I | R | D | O | E | Vi | Su | P | Fu | |||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Zhumanalieva Maksudakan | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | |||||||
| Omuralieva Gulsana | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | |||||||||||
| Abdykaarova Aigerim | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
C : Conceptualization I : Investigation Vi : Visualization M : Methodology R : Resources Su : Supervision So : Software D : Data Curation P : Project administration Va : Validation O : Writing - Original Draft Fu : Funding acquisition Fo : Formal analysis E : Writing - Review & Editing
Conflict of Interest Statement
The authors declare that there are no conflicts of interest regarding the publication of this article.
Informed Consent
Not applicable. This study analysed publicly available, de-identified data; no participants were recruited.
Ethical Approval
This study is a secondary analysis of de-identified ultrasound images released publicly under an open licence. No new data were collected from human participants, and ethical approval and informed consent were therefore not required.
Data Availability
The HC18 dataset analysed in this study is openly available under the CC BY 4.0 licence (Zenodo, doi:10.5281/zenodo.1327317). The ultrasound images shown in Figures 1 and 2 are from that dataset. The analysis code (Python, OpenCV) is available from the corresponding author on reasonable request.
Use of AI Tools
Generative AI tools were used to assist with code development, statistical analysis and drafting of the manuscript. The authors reviewed and verified all analyses, results and text and take full responsibility for the content of this article.
References
- [1] Salomon LJ, Alfirevic Z, Da Silva Costa F, Deter RL, Figueras F, Ghi T, et al. ISUOG Practice Guidelines: ultrasound assessment of fetal biometry and growth. Ultrasound Obstet Gynecol. 2019;53(6):715–723. doi.org/10.1002/uog.20272
- [2] American College of Obstetricians and Gynecologists. Committee Opinion No. 700: Methods for estimating the due date. Obstet Gynecol. 2017;129(5):e150–e154. doi.org/10.1097/AOG.0000000000002046
- [3] Salomon LJ, Alfirevic Z, Berghella V, Bilardo CM, Chalouhi GE, Da Silva Costa F, et al. ISUOG Practice Guidelines (updated): performance of the routine mid-trimester fetal ultrasound scan. Ultrasound Obstet Gynecol. 2022;59(6):840–856. doi.org/10.1002/uog.24888
- [4] Papageorghiou AT, Ohuma EO, Altman DG, Todros T, Cheikh Ismail L, Lambert A, et al. International standards for fetal growth based on serial ultrasound measurements: the Fetal Growth Longitudinal Study of the INTERGROWTH-21st Project. Lancet. 2014;384(9946):869–879. doi.org/10.1016/S0140-6736(14)61490-2
- [5] Chitty LS, Altman DG, Henderson A, Campbell S. Charts of fetal size: 2. Head measurements. Br J Obstet Gynaecol. 1994;101(1):35–43. doi.org/10.1111/j.1471-0528.1994.tb13007.x
- [6] Kurmanavicius J, Wright EM, Royston P, Wisser J, Huch R, Huch A, et al. Fetal ultrasound biometry: 1. Head reference values. Br J Obstet Gynaecol. 1999;106(2):126–135. doi.org/10.1111/j.1471-0528.1999.tb08212.x
- [7] Hadlock FP, Deter RL, Carpenter RJ, Park SK. Estimating fetal age: effect of head shape on BPD. AJR Am J Roentgenol. 1981;137(1):83–85. doi.org/10.2214/ajr.137.1.83
- [8] van den Heuvel TLA, de Bruijn D, de Korte CL, van Ginneken B. Automated measurement of fetal head circumference using 2D ultrasound images. PLoS One. 2018;13(8):e0200412. doi.org/10.1371/journal.pone.0200412
- [9] van den Heuvel TLA, de Bruijn D, de Korte CL, van Ginneken B. Automated measurement of fetal head circumference using 2D ultrasound images [dataset]. Zenodo; 2018. doi.org/10.5281/zenodo.1327317
- [10] Ramanujan S. Modular equations and approximations to π. Q J Pure Appl Math. 1914;45:350–372.
- [11] Hadlock FP, Deter RL, Harrist RB, Park SK. Estimating fetal age: computer-assisted analysis of multiple fetal growth parameters. Radiology. 1984;152(2):497–501. doi.org/10.1148/radiology.152.2.6739822
- [12] Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1(8476):307–310. doi.org/10.1016/S0140-6736(86)90837-8
- [13] Gray DL, Songster GS, Parvin CA, Crane JP. Cephalic index: a gestational age-dependent biometric parameter. Obstet Gynecol. 1989;74(4):600–603.
- [14] Sarris I, Ioannou C, Chamberlain P, Ohuma E, Roseman F, Hoch L, et al. Intra- and interobserver variability in fetal ultrasound measurements. Ultrasound Obstet Gynecol. 2012;39(3):266–273. doi.org/10.1002/uog.10082
Biographies of Authors
Zhumanalieva Maksudakan , is an Associate Professor in the Department of Clinical Disciplines at Jalal-Abad International University, Manas, Kyrgyzstan, where the author teaches and conducts research in clinical medicine. For this article, Zhumanalieva Maksudakan was responsible for conceptualization, obstetric interpretation of fetal biometry and gestational dating, supervision, writing – review and editing. Email: maksudabekievna68@gmail.com
Omuralieva Gulsana , is a Lecturer in the Department of Clinical Disciplines at Jalal-Abad International University, Manas, Kyrgyzstan, where the author teaches and conducts research in clinical medicine. For this article, Omuralieva Gulsana was responsible for methodology, formal analysis comparing perimeter formulae and cephalic index, writing – original draft. Email: samira132706@mail.ru
Abdykaarova Aigerim , is a Lecturer in the Department of Morphological Disciplines at Jalal-Abad International University, Manas, Kyrgyzstan, where the author teaches and conducts research in the morphological (basic medical) sciences. For this article, Abdykaarova Aigerim was responsible for data curation, measurement of head contours, investigation, visualization, writing – review and editing. Email: aygerim.talantovna@mail.ru