Supervised Contrastive Loss Helps Uncover More Robust Features for Photoacoustic Prostate Cancer Identification

DataThe data used in this study mainly comes from photoacoustic spectral analysis of biological tissues for prostate cancer (PCa) diagnosis. These tissue samples cover different individuals' prostate tissues, taking into account individual heterogeneity to mimic the diverse sample characteristics in real clinical settings. The data is utilized to evaluate the discriminative ability of different models in distinguishing between normal and cancerous tissues.CodeThe code is designed to implement multiple functions. Firstly, it constructs three distinct models, namely the CNN-based model, the supervised contrastive (SC) model, and the supervised contrastive loss adjust (SCL-adjust) model. It also includes the implementation of traditional feature extraction and common machine learning methods for comparison. Secondly, the code contains the logic for evaluating the performance of these models by calculating metrics like discriminative accuracy. Through running on the actual data, it can determine the advantages and disadvantages of each model under different conditions, such as in the presence of uniform or Gaussian noise. Overall, the data serves as the foundation and the code plays a crucial role in driving the analysis process and drawing key conclusions for improving the accuracy of PCa diagnosis.