Predicting HER2 overexpression in prostate cancer using machine learning: implications for personalized therapy.

Area of Science:

• Oncology
• Radiology
• Medical Imaging

Background:

• Human Epidermal Growth Factor Receptor 2 (HER2) overexpression is linked to advanced prostate cancer (PCa).
• Current research on HER2 in PCa primarily focuses on molecular pathology, with limited investigation into imaging biomarkers.
• There is a need for non-invasive methods to predict HER2 status in PCa.

Purpose of the Study:

• To develop a predictive model for HER2 overexpression in prostate cancer.
• To extract and utilize high-throughput radiomics features from MRI.
• To combine radiomics with clinical characteristics for enhanced prediction accuracy.

Main Methods:

• Retrospective analysis of 201 patients who underwent radical prostatectomy and HER2 immunohistochemistry.
• Extraction of multimodal radiomics features from T2-weighted imaging (T2WI) and apparent diffusion coefficient (ADC) maps.
• Development and evaluation of predictive models using LASSO regression and cross-validation, assessing discrimination and clinical utility.

Main Results:

• The combined model integrating ISUP Grade, PSA, and Radscore achieved an AUC of 0.841 in the test set.
• This model significantly outperformed the clinical model (AUC = 0.580) and showed modest improvement over the Radscore-only model (AUC = 0.838).
• The model demonstrated consistent discriminatory power with 0.78 accuracy, 0.77 sensitivity, and 0.79 specificity, alongside strong clinical applicability.

Conclusions:

• A combined radiomics and clinical model effectively predicts HER2 overexpression in prostate cancer.
• This predictive model has the potential to guide personalized treatment strategies for HER2-overexpressing PCa.
• Integrating imaging biomarkers offers a promising avenue for non-invasive assessment of HER2 status in PCa.