AI-Assisted Tumor Boundary Delineation via Targeted Ultrasmall Iron Oxide Nanoprobe for High-Contrast HER2-Positive Tumor Imaging.
Authors
Affiliations (2)
Affiliations (2)
- Department of Biomedical Engineering, College of Engineering, Shantou University, Shantou, People's Republic of China.
- Department of Radiology, Shanghai General Hospital, Shanghai Jiaotong University School of Medicine, Shanghai, People's Republic of China.
Abstract
Breast cancer continues to be a leading cause of cancer-related mortality in women globally, where precise diagnosis and clear tumor demarcation are critical for effective treatment. Herein, we developed a strategic platform that combines a novel ultrasensitive magnetic resonance (MR) contrast agent with deep learning to significantly enhance the tumor-to-normal ratio (TNR). We designed ultrasmall iron oxide nanoparticles (USIO NPs) conjugated with trastuzumab (Tmab) for targeted MR imaging of HER2-positive breast cancer. The USIO@Tmab nanoprobe demonstrated excellent HER2 specificity and pH-responsive activation. The relaxivity of the nanoprobe shifted from a low T<sub>1</sub>-weighted intensity (r<sub>1</sub> = 1.43 mM<sup>-</sup> <sup>1</sup>s<sup>-</sup> <sup>1</sup>) under physiological conditions to an enhanced value (r<sub>1</sub> = 4.07 mM<sup>-</sup> <sup>1</sup>s<sup>-</sup> <sup>1</sup>) in the acidic tumor microenvironment due to the detachment of Tmab protein. Additionally, we employed the 3D nnU-Net deep learning framework as a post-processing visualization aid to enhance tumor boundary detection via image fusion, rather than to amplify the underlying MRI signal. This approach yielded high segmentation accuracy, with an Intersection-over-Union (IoU) of 0.88 and a Dice coefficient of 0.93. This strategy provided an additional 2.59-fold increase in TNR and enabled the reconstruction of three-dimensional (3D) tumor models, offering clinicians an intuitive visualization of tumor structure for precise diagnosis and surgical guidance.