Prostate cancer is increasingly prevalent due to the aging population, imposing significant economic and psychological burdens [1,2]. According to the National Comprehensive Cancer Network (NCCN) guidelines (v4.2023), mutations in homologous recombination repair (HRR) genes are a major cause of homologous recombination deficiency (HRD) in prostate cancer. These HRR gene mutations are identified as alterations in genes such as BRCA1, BRCA2, ATM, BARD1, BRIP1, CDK12, CHEK1, CHEK2, FANCL, PALB2, RAD51B, RAD51C, RAD51D, and RAD54L [3]. HRR genes mutations are present in up to 30 % of patients with metastatic castration-resistant prostate cancer (mCRPC), and these mutations are linked to an early onset of disease, aggressive tumor behavior, high recurrence rates, and poor prognosis [4,5]. In terms of treatment, several targeted poly ADP-ribose polymerase (PARP) inhibitors, such as Olaparib, Niraparib, and Talazoparib, are available for patients with HRR gene mutations. These drugs exert their effects by inducing synthetic lethality, inhibiting DNA repair mechanisms in tumor cells, and thereby enhancing the efficacy of other antitumor agents. Clinical evidence suggests that Olaparib significantly prolongs progression-free survival (PFS) in mCRPC patients with HRR mutations, and Luca Parib has demonstrated a 44 % objective response rate (ORR) in patients with BRCA1/2 mutations, underscoring the potential of PARP inhibitors to improve both PFS and overall survival (OS) in this patient population [3,6,7]. Given the clinical implications of HRR genes mutations in PCa, non-invasive assessment of these mutations could provide valuable insights for personalized treatment and prognostication [8].

However, conventional genetic testing methods, while informative, are invasive, costly, and often operationally complex. As such, there is a pressing need for a non-invasive, cost-effective, and straightforward tool for identifying HRR mutations. Multiparametric magnetic resonance imaging (mpMRI) is capable of revealing detailed characteristics of prostate cancer lesions and is relatively affordable compared to other diagnostic methods. Recent studies have explored the potential of mpMRI as a surrogate for gene mutations in prostate cancer. For example, McCann et al. identified a significant correlation between the expression of phosphatase and tensin homolog (PTEN) and dynamic contrast-enhanced (DCE)-MRI features in the peripheral zone of the prostate, as well as with Gleason scores [9]. Additionally, our previous work demonstrated that mpMRI-based models, in combination with Gleason scores, were effective in predicting TP53 mutations [10]. Su et al. demonstrated that MRI-based radiomics can noninvasively predict HRD status in triple-negative breast cancer, achieving an AUC of 0.739 [11]. Recent studies demonstrated that deep learning artificial intelligence was able to predict HRD status and platinum response directly from conventional histologic slides [12]. This breakthrough established the feasibility of using image-based computational methods to assess molecular alterations at the genetic level. Subsequent work explore the potential of extending this predictive capability to non-invasive imaging modalities such as MRI. However, there is currently limited research on predicting homologous recombination repair (HRR) gene mutations in prostate cancer.

The objective of this study is to investigate the relationship between radiomics features derived from mpMRI and HRR genes mutations in prostate cancer. After identifying non-invasive imaging biomarkers that correlate with HRR genes mutations, this research aims to develop a predictive model with these biomarkers and clinicopathological data, to assist clinicians in identifying patients likely to benefit from targeted therapies, thereby improving the prognosis and survival outcomes of prostate patients.