PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
September 23, 2025European Journal of Endocrinology0 citationsOpen Access

OC43 - Radiogenomics pilot study in adrenocortical carcinoma: assessing the relationship between genetic background and computerized tomography texture

View Full Paper
LTLorenzo TucciGVG. VaraALAntonio De Leo

Key Points

  • Radiomics predicted pathogenic mutations in adrenocortical carcinoma with a model F1-score of 0.70.
  • Key findings showed 56.5% of patients had pathogenic/likely pathogenic variants, including in β-catenin pathway genes.
  • Utilizing deep learning and logistic regression models, the study compares mutation predictions in ACC.
  • Further validation is needed on a larger cohort to enhance current findings and establish clinical relevance.

Abstract

Abstract Background/Introduction Some somatic mutations have prognostic relevance in adrenocortical carcinoma (ACC). Radiomics has been explored for mutation prediction but not in ACC. Purpose To predict the presence of key pathogenic gene variants in ACC with radiomics. 2. To compare the ability of radiomics vs β-catenin nuclear staining (as reported from two previous papers) to predict mutations in β-catenin pathway genes (CTNNB1, ZNRF3, APC). Methods We enrolled 46 patients with ACC (32 females) from Birmingham (UK) or Bologna (Italy). Targeted Next-Generation Sequencing was performed on DNA extracted from formalin-fixed paraffin-embed samples using a customised panel with 10 ACC-specific genes. Radiomics was performed on portal-phase computerized tomography. We formulated deep learning model (DLM) to predict overall presence of variants and logistic regression models (LRM) to predict mutations in β-catenin pathway genes, tumour suppressor genes (TP53, RB1, CDK4), chromatin remodeling genes (MEN1, TERT), NF1 and ATM. All models were based on radiomics features. Only pathogenic/likely pathogenic variants (P/LPv) and variants of unknown significance (VUS) were included in the analysis. Results P/LPv and VUS were detected in 26 cases (56.5%), including β-catenin pathway genes (n=9 (19.6%): CTNNB1=7 (15.2%), ZNRF3=2 (4.3%), APC=1 (2.2%)), tumour suppressor genes (n=14 (30.4%): TP53 =12 (26.1%), RB1=2 (4.3%)), MEN1=4 (8.7%), TERT=3 (6.5%), NF1 (n=8 (17.4%)). DLM predicted overall presence of variants with F1-score=0.70. LRM predicted P/LPv and VUS in β-catenin pathway genes overall (F1-score=0.6), and in CTNNB1 (F1-score=0.55) TP53 (F1- score=0.64), chromatin remodeling genes (F1-score=0.66), and MEN1 (F1-score=0.5), specifically. In previous publications, β-catenin nuclear staining showed F1- score=0.52, F1-score=0.42 in predicting CTNNB1 mutations, and F1-score=0.73, F1- score=0.45 in predicting mutations in β-catenin pathway genes. Conclusion(s) Radiomics could be useful for early recognition of P/LPv and VUS in ACC and could be an alternative to β-catenin nuclear staining. Further studies on wider cohort are required to improve current findings.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Tucci et al. (2025) studied this question.

synapsesocial.com/papers/68d4724f31b076d99fa6abechttps://doi.org/10.1093/ejendo/lvaf168.080
Ask AI
Helpful
Bookmark
Share
View Full Paper