China
Africa
Explore chapters and articles related to this topic
The essence of R in head and neck cancer
Published in Ruijiang Li, Lei Xing, Sandy Napel, Daniel L. Rubin, Radiomics and Radiogenomics, 2019
Hesham Elhalawani, Arvind Rao, Clifton D. Fuller
Afterward, investigators sought to include more clinically meaningful oncologic outcome endpoints other than overall survival. The Medical Image Computing and Computer Assisted Intervention Society (MICCAI) 2016 radiomics challenge identified “local tumor control” as a testable clinically oriented hypothesis with a binary endpoint that can be used for radiomics challenges using an existing dataset of 288 RT-treated OPC patients (Elhalawani et al. 2017, 2018b). A series of studies ensued where radiomics was more geared toward “predicting” the outcomes of RT in the definitive setting aiming at pre-emptively risk-stratifying the patients in terms of their subsequent local tumor control. A 3-feature radiomics signature (L: low-pass filter; H: high-pass filter [HHH] large size high gray-level emphasis, LLL sum entropy, and L: low-pass filter; H: high-pass filter [LLH] difference variance) was significantly associated with local control both in training (n = 93) and validation (n = 56) HNC cohorts, respectively (Bogowicz et al. 2017). Similarly, an ensemble of three histogram and four textural features composed a predictive CE-CT signature of local tumor recurrence following RT in a cohort of 62 patients with HNC.
Cardiovascular Health Informatics Computing Powered by Unobtrusive Sensing Computing, Medical Image Computing, and Information Fusion Analysis
Published in Ayman El-Baz, Jasjit S. Suri, Cardiovascular Imaging and Image Analysis, 2018
Chengjin Yu, Xiuquan Du, Yanping Zhang, Heye Zhang
This section focuses on computing in cardiovascular health informatics, which includes sensing computing, medical image computing, and information fusion analysis. In the past, doctors and researchers collected medical data by sensing and imaging. Limited time and effort can be put into the analysis of the data. With the enhancement of computing power and the miniaturization of electronic devices, the ideal of continuous unobtrusive sensing becomes reality. It is very desirable for us to develop computer-aided diagnosis systems and even fully operative computer diagnosis systems when we face big data. An example is given in Figure 7.1, which describes the evolution of the electrocardiogram (ECG) device. Electronic devices evolved from water buckets and bulky vacuum tubes to discrete transistors-based machines. Their size also evolved from desktop sized to small, wearable equipment [5]. A clear trend is that they will become smaller, lighter, and more comfortable to wear.
Bias in machine learning for computer-assisted surgery and medical image processing
Published in Computer Assisted Surgery, 2022
John S. H. Baxter, Pierre Jannin
However, train/test leakage is more insidious and some studies have shown it to have a large effect on the medical imaging and computer-assisted surgery literature, thus going unnoticed in the review process [10,11] both investigate this in two different applications, for example). Simply put, train/test leakage is when information from the test set is provided in training time, although the simplicity of this formula denies the complexity of the problem as different data points themselves are correlated and provide information about each other. To give an example, several medical image computing datasets include multiple images of the same patient across different time points or even just two-dimensional slices from the same three-dimensional volume. (One reason why these types of errors are so insidious is the complexity of some evaluation methods, such as cross-validation and nested cross-validation, where the difference between a correct and a leaking implementation can be difficult to detect for both authors and reviewers.) There are obvious correlations here that could be leaked if images from the same patient found themselves in both the training and the evaluation datasets at the same time. However, this is also the case with particular hospital centers. Should multi-center experiments require all of the images from each center to fall on the same side of the training/evaluation divide?
Validation of a coupled electromagnetic and thermal model for estimating temperatures during magnetic nanoparticle hyperthermia
Published in International Journal of Hyperthermia, 2021
Sri Kamal Kandala, Anirudh Sharma, Sahar Mirpour, Eleni Liapi, Robert Ivkov, Anilchandra Attaluri
The CT images of rabbit were imported into an open-source software, 3DSlicer [60], for visualization and medical image computing. Automatic segmentation does not yield a desirable result due to the overlapping gray level values of the soft tissue organs. 3DSlicer was used to manually segment and differentiate the liver from the other organs using the gray level value and apply smoothing algorithms to the region of interest. Manual segmentation was performed to differentiate the liver from other organs by choosing the lower and higher threshold gray level value. Segmented files were converted into a 3D surface geometry CAD (.stl) format. The meshing of the liver CAD geometry was performed using a Laplacian mesh smoothing method in MeshLab® [61]. Several iterations of mesh smoothing were carried out to ensure a well smoothened geometry. Surface treatment and defeaturing were performed to smooth sharp edges and artifacts obtained in reconstructed geometries to ensure compliance of the meshed geometry (.stl file) with finite element simulation software. The workflow of building 3D models from CT scan images is shown in Figure 3.
A review of haptic simulator for oral and maxillofacial surgery based on virtual reality
Published in Expert Review of Medical Devices, 2018
Over the past years, with the development of computer technologies, virtual reality (VR) and haptic feedback have been becoming hot topics in the field of computer-assisted surgery, bringing great changes to the traditional medical training [4,5]. Haptic-based surgical simulation is an interdisciplinary area with the integration of computer graphics, medical image computing, mathematics, biomechanics, surgery, and so on. The basic architecture is shown in Figure 1. With the use of haptic feedback and visual immersion technology, the fidelity of visual and tactile fusion of virtual environment can be guaranteed. It can be carried out on the patient-specific models, which are conducive for the doctor to observe the lesion, and to facilitate preoperative surgical planning and training [6] as well as intraoperative assistance. It is also conducive for medical students or novice surgeons to simulate the surgical operation, and the VR systems benefit users in terms of the understanding of the surgical procedures and enhancement of the surgical operation skills. Haptic surgery simulators based on VR in other medical training fields [7–12] have proved to be safe, accurate, cost-effective, and repeatable. Virtual OMFS simulators have been developed for oral implantology [13], trauma cutting [14], orthognathic surgery [15], and so on, providing an alternative to traditional training methods. Another significant advantage of VR-based simulator is that it can record data automatically while simulating and evaluating operator’s performance objectively.