In this manuscript, we report on a well-characterized, gestational age-matched cohort of in total n = 66 prenatal MRI reports for different indications, where one third of the population reported PAE. Fetal liver volumes did not differ between fetuses with and without PAE—even when adjusted for gestational age. Also, signal intensities of the fetal liver did not differ between the two groups. To identify miniscule alterations of the liver, a texture-based Radiomics classification model was created to potentially identify PAE-related changes. However, the designed model could not differentiate between patients with or without PAE. Hence, we assume that PAE causes no relevant effect on the fetal hepatic parenchyma or organ maturation detectable by classical MRI and/or radiomics.

Although there is no safe level of alcohol consumption during pregnancy, prenatal alcohol exposure is still a common risk factor for fetal developmental delay—especially in Europe, where a PAE rate of 25% was reported in some countries [5, 9]. In the United States of America and in Europe, the prevalence of FASD is high, with an estimated 2–5 fetuses per 100 births [4, 26]. While the role of PAE in the development of FASD is well established and fetal MRI is capable of detecting alterations of the fetal brain after alcohol exposure [16, 27, 28], its effect on development of the fetal liver is still unclear. PAE has been associated with low weight, insulin resistance, growth retardation after birth, and the development of diabetes mellitus later in life [20, 29, 30]. Together, these factors may further contribute to metabolic dysfunction-associated steatotic liver disease (MASLD) in the adolescent and adult patient [21, 31]. Furthermore, an association of PAE and the development of some childhood cancers like neuroblastoma, leukemia, and brain tumors has been described [32, 33]. While the role of PAE in the development of FASD is well established and fetal MRI is capable of detecting alterations of the fetal brain after alcohol exposure [16, 27, 28], data regarding the effect of PAE on the fetal liver are lacking. Besides superficial description of changes in imaging modalities, radiomics might allow a deeper assessment and quantification of imaging data and detection of alterations otherwise inaccessible to the human eye [11]. Radiomics has already been applied in fetal MRI of the lung, where a radiomics-based approach allowed evaluation of lung maturity [17] or in fetal MRI and radiomics-based assessment of the placenta to predict fetal growth restriction [34]. Regarding liver conditions, application of radiomics on MRI-derived data allows the identification of alcohol exposure in adult humans and rats [35]. Radiomics also allows assessment of liver fibrosis and an estimation of etiology in adult patients [12, 13], but data regarding its use in fetal liver MRI are lacking.

In our study, no significant effect of PAE on the fetal liver was discernable in radiological analysis, as well as via texture-based radiomics analysis of the extracted imaging data. The observed AUC values of the random forest or logistic regression models were both slightly below 0.5 (0.43 and 0.45, respectively, Fig. 4), which might be suggestive of inverse or confounding signals in the dataset. Data acquisition was performed carefully, and no potentially affecting covariates were identified. Furthermore, for both models, the confidence intervals spanned across 0.5. Hence, we assume that the AUCs below 0.5 are mere chance and not due to a relevant bias in the dataset. While it has been shown in adults that microstructural remodeling in the liver often produces meso-/macro-scale imaging correlates—i.e., changes in texture, intensity distributions, and spatial heterogeneity—that radiomics can quantify, based on our data, we cannot exclude that PAE produces metabolic changes to the liver that cannot be captured by radiomics [36,37,38]. Further research, in particular on laboratory parameters of newborns with PAE and/or post-mortem studies, might help to elucidate the potential mechanisms.

Classification of prenatal alcohol exposure status based on fetal liver magnetic resonance imaging radiomics features. Based on liver MRI radiomics features, neither random forest nor L2-penalized logistic regression models were able to discriminate between fetuses with and without prenatal alcohol exposure

While no macroscopic alterations of the fetal liver were observed in the PAE cohort, the detrimental effects of alcohol exposure during pregnancy are well-known, and the World Health Organization recommends complete cessation of alcohol consumption during pregnancy as no safe level is known [39, 40]. Children with FAS exhibit craniofacial (thin upper lip, smooth philtrum, short palpebral fissure) and cerebral alterations (decreases in gray matter thickness, volume reduction, decreased cortical gyrification) post partum [4]. The individual indications for fetal MRI in our cohort are presented in Supplementary Table 1, and while the cohort size is relatively small and the reported outcomes heterogeneous, alterations of the head and face (microcephaly, midface hypoplasia, exophthalmos, cleft lip and palate) were exclusive to the PAE cohort. Furthermore, during pregnancy, the placenta assumes metabolic pathways usually domains of the liver, and thus, potentially sparing the fetal liver of the detrimental effects of alcohol, which might explain the absence of detectable hepatic alterations [41, 42]. In line with this, placental aberrations have been described in mothers with alcohol consumption during pregnancy [16, 43, and Stuempflen et al (2026), unpublished]. In our cohort, alterations of the placenta were exclusive to the PAE cohort as seen in Supplementary Table 2.

Despite the absence of structural MRI correlates, postnatal biochemical alterations were observed in some PAE-exposed newborns, suggesting that metabolic changes—physiological or pathological—may precede or occur independently of macrostructural changes detectable by current MRI techniques. Post-partum laboratory data were available in four patients of the PAE cohort. Transaminases, like aspartate transaminase (AST) and alanine aminotransferase (ALT), are enzymes inside the hepatocytes’ cytoplasm or mitochondria, and elevated serum levels are indications of hepatic inflammation and/or necrosis [44]. Although broadly available biomarkers for liver damage, no universally accepted normal range for AST or ALT has been established in adult patients due to globally varying kits and cut-offs; however, 40 U/L is commonly referred to as “upper limit of normal” [45,46,47]. Similar levels have been described in newborns, with usually AST higher than ALT and a mild effect of alcohol consumption on AST levels [48, 49]. In our cohort, despite no evident changes on MRI or MRI-based radiomics, three out of four fetuses with available post-partum laboratory data had transaminases above 40 U/L. Elevated AST/ALT values tended to decline spontaneously after birth. Another hepatic enzyme, the gamma-Glutamyl-Transferase (gGT), was also elevated above the level of adult patients in three out of four fetuses. gGT is associated with alcohol consumption in adults; however, also associated with the induction of hepatic activity in the newborn, and increasing levels after birth are expected [50]. One fetus of a mother with reported binge drinking had elevated gGT at birth, which declined in the following days. Alkaline phosphatase (ALP), a hepatic enzyme associated with bile flow and cholestasis, was slightly elevated in one fetus of a mother who reported binge drinking episodes. However, ALP is also involved in bone metabolism, and high blood levels can be expected in newborns [51]. See Supplementary Fig. 2.

Our study has limitations we want to address. As fetal MRI is not a widely available procedure and standardized assessment of alcohol consumption are rarely available with imaging reports, the sample size is relatively small. However, the cohort and sub-cohorts were homogeneous without apparent differences in relevant characteristics. The confidence intervals of the compared groups were predominantly overlapping, supporting the robustness of the data and the described lack of observable biological effect on the fetal liver on MRI. Although alcohol consumption was assessed by standardized questionnaires, the data were self-reported. No other methods to evaluate alcohol consumption, like blood-based tests, were used. The small sample size is especially apparent in the small sub-group of mothers who reported binge drinking episodes during pregnancy (n = 5 of n = 21 mothers with alcohol exposure). In this sub-group, we found smaller organ volumes of fetal livers compared to fetuses without PAE; however, the finding did not maintain its statistical significance when adjusted for gestational age. This advises for careful interpretation as the effect might be unstable due to the small sub-sample size and highlights the need for further investigation in larger cohorts, as despite the negative data in our manuscript, an effect from high-level exposure could exist. While quantitative MRI approaches might be superior in depicting the biochemical and microstructural processes of liver injury, image acquisition times and fetal motion limit their use in fetal MRI in vivo, and they are not part of standardized fetal MRI acquisition protocols [52, 53]. Post-partum laboratory data were not available in all patients, as some patients were external referrals and follow-up was conducted at a different center. Stuempflen et al assessed PAE-associated alterations of the fetal brain in an overlapping cohort and found significant changes to the corpus callosum and periventricular zone in alcohol-exposed fetuses [16]. The reported effects are potential forebearers of postnatal delays in neuro-development, compatible with FASD [54]. In our cohort, brain alterations detectable on fetal MRI were exclusive to the PAE cohort and are reported in Supplementary Table 2.

In summary, fetal MRI is capable of a detailed assessment of fetal organs that complements ultrasound in many ways. Transferring these imaging reports into radiomics data may allow an even more detailed and deeper analysis. In our manuscript, no direct effect of alcohol exposure on fetal liver maturation or MR-morphology was detectable. Deeper analysis through a radiomics-based approach did not identify any potentially alcohol-related changes in the fetal liver. For the clinical radiologist and obstetrician/gynecologist, alterations of liver morphology on imaging studies should prompt the search for differential diagnoses other than alcohol-related complications.