Deep Learning Techniques for Radioisotope Identification in Complex Gamma-Ray Spectra: A Comprehensive Survey

: Fast nuclide identification with increased accuracy rates is an important demand in nuclear applications. So, effective spectrum identification techniques are established for low-resolution NaI(TI) scintillator and Si(Li) detector spectra. The goal of recent research is to identify X-ray and gamma complex spectra. Conventional spectrum analysis techniques often struggle with overlapping spectral peaks, background radiation, detector noise, and mixed-isotope settings, which limits their efficacy in real-world applications. This overview covers radioisotope identification methods from peak-based and statistics to machine learning algorithms and deep learning models for automated gamma-ray spectrum categorization. It also evaluates previous methods' accuracy, computing efficiency, generalization, and deployment practicality. The paper highlights explainable artificial intelligence, transfer learning, self-supervised learning, and lightweight edge-deployable models, as well as unsolved issues including limited annotated datasets and cross-detector adaptability. The findings can help researchers and practitioners develop reliable and intelligent radioisotope identification systems for nuclear threat detection, border security, radioactive waste management, emergency response, and nuclear medicine, including PET, SPECT, and radiopharmaceutical quality assurance.