A polyline searching-driven evolutionary AI for disease detection of medical imaging data

Objective Accurate Ultrasound (US) prostate cancer (PCa) segmentation images hold significant value for organ interventional guidance and clinical disease diagnosis. However, this task still poses substantial challenges. The main obstacles include blurred or incomplete boundaries separating PCa from adjacent soft tissues, shadow artifacts inherent to ultrasound imaging, and drastic inter-patient variations in organ morphological shapes. Approach To address these issues, our method introduces a novel coarse-to-fine optimization framework to overcome these challenges on ultrasound data, which comprises four main stages: 1) initial region of interest (ROI) localization using a deep learning model; 2) vertex sequence determination through a principal curve-based polyline search; 3) optimal initialization of the backpropagation neural network via an enhanced quantum evolutionary algorithm; 4) defining the ROI boundary with a mathematical model based on neural network parameters. Main results Our method outperforms current algorithms in ultrasound PCa data, achieving mean Dice similarity coefficient (DSC), Jaccard similarity coefficient (OMG), and accuracy (ACC) of 83.6±3.1%, 71.8±2.5%, and 83.5±3.1%, respectively. Meanwhile, our method achieved the best segmentation capability using public Digital Database of Thyroid Ultrasound Images (DDTI) dataset for external evaluation. Significance This paper proposes an intelligent algorithm for ultrasound PCa segmentation, where accurate ultrasound PCa segmentation is critical for protecting vulnerable anatomical structures during clinical procedures. Our method is expected to improve diagnostic accuracy and facilitate better treatment efficacy.