Using an Inspired Quantum-Classical Sequential Neural Network to Segment Liver Tumor Images Accurately
2708-9967
Abstract
Medical analytics based on multimodal intelligence are essential to contemporary healthcare, especially in the diagnosis and detection of liver cancer. However, low contrast, varied textures, and irregular tumor forms in medical imaging continue to make precise liver tumor segmentation difficult. These intricate features are frequently difficult for traditional intensity-based and classical deep learning techniques to adequately capture. This study suggests a unique Inspired Quantum-Classical Sequential Neural Network (IQCSNN) for accurate liver tumor picture segmentation in order to overcome these drawbacks. The suggested method is new because it combines classical sequential neural networks with quantum computational concepts to improve segmentation accuracy and feature representation. To enhance picture quality, preprocessing methods such as normalization, noise reduction, and standardization are used to liver CT images obtained via Kaggle. Following preprocessing, the data is sent into the IQCSNN model, which uses quantum-classical feature extraction techniques to identify both local and global patterns. The Inspired quantum layer enhances the ability to distinguish between tumor and non-tumor tissues, while the sequential neural network component facilitates efficient learning of spatial connections. The usefulness and potential of the suggested strategy for advanced medical image analysis are highlighted by experimental findings, which show that it produces better segmentation performance than traditional methods.