Low-dose CT Image Quality Assessment Based On PCJO

Computed tomography(CT)has been widely applied in the field of medical domain nowadays.However,the high risk of radiation damage along with CT scanning has received serious concern,which brings low-dose CT(LDCT)great importance.LDCT lowers the dose of CT scanning by means of cutting down the tube current or the tube voltage,but at the cost of more noise and artifacts,which reduces the quality of LDCT images.Therefore,LDCT reconstruction algorithms and relevant processing algorithms with excellent performance are necessary in order to guarantee the quality of reconstructed images.Accurate algorithms of medical image quality evaluation are also required in order to compare and assess reconstruction algorithms.Although subjective assessment is reliable,there exists problems of long working time?high human burden?poor real-time and prone to be influenced by subjective and objective factors in the subjective assessment.Therefore,accurate objective evaluation is necessary to conduct auxiliary assessment so as to reduce the burden of doctors and cut down the manpower cost.Modern objective evaluation algorithms can be divided into two categories:the assessment based on human visual system(HVS)and the assessment based on the model observer(MO).These two objective assessment methods usually incorporate with the task-based assessment method that is now commonly applied in the medical image quality evaluation.In order to testify the adaptation of PCJO to the LDCT field and explore the effect of auxiliary assessment,we propose a new LDCT objective image quality assessment based on PCJO.This method can be separated into two modules:training module and testing module.The training module mainly utilizes the joint detection and estimation theory and the channelized method based on sub-band decomposition characteristic of HVS.The testing module gets the testing blocks and then conducts detection-localization task with visual difference predictor(VDP).Specific experiments include objective assessment experiments based on FBP?TV and FCR reconstruction image sets and the subjective assessment experiment for comparison.Preliminary experiments have revealed that as a classical kind of the anthropomorphic model observer,PCJO shows outstanding performance in the medical image assessment based on LDCT image sets and the target task of multiple-lesion detecting and locating.On the condition that accurate VDP threshold and channels are selected,PCJO can approach the performance of subjective assessment.Afterwards,we change our FOM from the JAFROC1 to the wJAFROCl that assigns every lesion of each image the weight in order to explore wJAFROCl's task performance of lesion detection-localization.We conduct quality assessment on these three reconstructed image sets using wJAFROC1.Experimental results reveal that results can be achieved consistent with results of JAFROC1.Meanwhile,analyzed from the evaluation results of FOM and VDP threshold,we can verify the reconstructive performance of these three LDCT reconstruction algorithms.Experimental results show that FCR attains the best reconstructed image quality,followed by TV the second and FBP the third.