A data-driven maximum likelihood classification for nanoparticle agent identification in photon-counting CT
| Autor: | Sumin Baek, Okkyun Lee |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Likelihood Functions
Photons Radiological and Ultrasound Technology Phantoms Imaging business.industry Computer science Detector Metal Nanoparticles Estimator Pattern recognition Photon counting k-nearest neighbors algorithm Background noise Robustness (computer science) Radiology Nuclear Medicine and imaging Gold Sensitivity (control systems) Artificial intelligence Tomography X-Ray Computed business Energy (signal processing) |
| Zdroj: | Physics in Medicine & Biology. 66:145004 |
| ISSN: | 1361-6560 0031-9155 |
| Popis: | The nanoparticle agent, combined with a targeting factor reacting with lesions, enables specific CT imaging. Thus, the identification of the nanoparticle agents has the potential to improve clinical diagnosis. Thanks to the energy sensitivity of the photon-counting detector (PCD), it can exploit the K-edge of the nanoparticle agents in the clinical x-ray energy range to identify the agents. In this paper, we propose a novel data-driven approach for nanoparticle agent identification using the PCD. We generate two sets of training data consisting of PCD measurements from calibration phantoms, one in the presence of nanoparticle agent and the other in the absence of the agent. For a given sinogram of PCD counts, the proposed method calculates the normalized log-likelihood sinogram for each class (class 1: with the agent, class 2: without the agent) using theKnearest neighbors (KNN) estimator, backproject the sinograms, and compare the backprojection images to identify the agent. We also proved that the proposed algorithm is equivalent to the maximum likelihood-based classification. We studied the robustness of dose reduction with gold nanoparticles as the K-edge contrast media and demonstrated that the proposed method identifies targets with different concentrations of the agents without background noise. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|