Background

Spectral CT is an imaging modality that allows for tissue‑specific texture differentiation, along with anatomical information provided by conventional CT. However, current methods of image reconstruction for spectral CT fail to fully take advantage of this potential, and result in degradation of texture quality. Texture is a crucial feature used to distinguish healthy tissue, benign lesions, and malignant lesions. Conventional CT and current spectral CT methods often lack sufficient texture quality required for diagnosis, resulting in misdiagnosis, unnecessary invasive procedures, missed and untreated lesions, or increased radiation dosage due to further imaging.

Technology

This technology utilizes artificial intelligence to improve spectral CT reconstruction tissue‑specific texture quality and assist in computer‑aided lesion diagnosis, improving the capabilities of spectral CT scanning for lesion diagnosis. The reconstruction of the spectral CT images is achieved by considering the structural similarity among the energy spectral channels, the tissue-specific textures in each energy channel, and spectral data statistics. The ultimate goal is to enhance the tissue-specific textures in each energy channel by using the structural similarity among all energy channels, the tissue-specific textures in each energy channel, and the data statistics, all in a single equation. Using the improved images, texture-based computer-aided diagnosis (CADx) is able to fully utilize the energy spectral tissue-specific textures for lesion diagnosis. The addition of this energy spectral tissue-specific texture CADx can maximize the potential benefit of CT scanning for medical diagnosis.

Advantages

Improved texture quality for spectral CT images - Increased differentiation of tissue‑specific textures in CT images - More accurate texture‑based computer‑aided diagnosis of lesions

Application

Medical imaging (Hospitals, Radiology clinics, Cancer centers) - Research