| Summary: | 碩士 === 國立中正大學 === 資訊管理系醫療資訊管理研究所 === 107 === Heavy workload and long hours of continues work will cause overwork. It is a threat to the radiologist who is sitting in front of the desk and typing the image report. Therefore, if a decision-making system can be introduced to assist radiologist writing report. It will reduce the burden on radiologists for image interpretation. and also assists radiologists to figure out the lesion rapidly. It allows patients to further treatment to improve survival. The research use liver MRI image as a material. and also use machine learning technology to figure out the tumor in the image. And then develop it into a predictive model to assist in the interpretation of tumor images.
We collected 571 patients of liver MRI image. In addition, the images are pre-processed and further extracted. Finally, the extracted features are analyzed using a support vector machine classification technique to construct a predictive model for predicting tumors. Finally, the five models is evaluated by six criteria. In terms of precision, the four groups of predictive models predict the highest tumor value(0.685), Indicating that the predictive models have a high predictive rate on T2 MRI images. We also figure out the predictive models have the highest F value on +C MRI images which means we need to use contrast for assisting the images to find out the tumor. In future it will induced artificial intelligence to improved the probability of tumor discovery. And also ust the technique to other cancer images to assist radiologists to figure out the tumors. It allow patients to receive treatment early, and also improving their survival rate and medical quality.
|
|---|
