An integrated digital twin framework for tumour progression prediction and dynamic monitoring of 3D bioprinted breast tissue

Tumour progression needs to be accurately analysed in complex structures of the breast tissue that requires sophisticated intelligent frameworks that can continuously classify, predict and monitor. In this research, a computational digital twin-based framework is presented for tumours with a 3D model based on breast tissue-oriented image of bioprinted breast tissue constructs is presented. Radiomic features of texture, intensity and structural heterogeneity are derived on the segmented regions and the most informative features of each are identified with a Random Forest-based feature selection method. The proposed Bioprinted Tissue-Aware Latent Feature Classification Network (BTLFC-Net) is used to classify tumours, which is trained to learn discriminative latent tissue representations. Latent Feature-Driven Tumour Prediction Network (LFTP-Net) is utilised to predict the evolution of tumour with time whereas Digital Tumour Multimodal Monitoring Network (DTM-Net) allows continuous assessment of tumour through continuous updating of a virtual tumour replica. Experimental findings reveal that the proposed framework attains high classification and effective tracking of tumour progression as compared to the current methods. Tissue-aware learning combined with digital twin modelling allows predicting and monitoring tumours accurately, thus helping to make individual and proactive clinical decisions.