Invented by Katelyn Rose Nye, Gireesha Rao, General Electric Co
The General Electric Co invention works as followsApplications, systems, as well as methods of delivering point-of-care alerts for radiological findings have been disclosed. An example imaging apparatus comprises an image data storage, an image quality controler, and a learning network. An example image store stores image data that has been acquired with the imaging apparatus. The image quality checker evaluates image data stored in the image storage and compares it to an image quality measurement. The trained learning network will process the image data in order to identify a clinical problem. This will trigger an alert at the imaging apparatus so that a healthcare practitioner is informed and prompts a response with regard to the patient.
Background for Systems and methods for delivering point-of-care alerts for radiological findings
Healthcare facilities such as hospitals, clinics and doctors’ offices face many economic, technological, and administrative challenges in order to provide high-quality care for patients. The emerging accreditation to control and standardize radiation exposure dose usage within a healthcare organization, as well as a lack of skilled staff and equipment, are all factors that can hinder the effective management of imaging and information systems used for diagnosis, treatment, and examination of patients.
Healthcare provider consolidations create geographicly dispersed hospital networks, in which physical contact with the systems is too expensive. Referring physicians also want better collaboration and direct access to data. Physicians are often overwhelmed with data and have less time. They are also more patient-oriented and eager to get help.
Healthcare provider (e.g. x-ray technologist or doctor, nurse, etc.). Tasks such as image processing, analysis, quality control/quality assurance, and the like are difficult and time-consuming tasks that humans cannot do alone.
Certain illustrations provide apparatus, systems and methods to improve image quality control, image processing and identification of findings in data. They also generate notification at or near the point of care for patients.
Some examples include an imaging apparatus that includes an image data storage, image quality checker and a trained learning system. An example image store stores image data that has been acquired with the imaging apparatus. The image quality checker evaluates image data stored in the image storage and compares it to an image quality measurement. The trained learning network will process the image data in order to identify a clinical problem. This will trigger an alert at the imaging apparatus so that a healthcare practitioner is informed and prompts a response with regard to the patient.
Certain example instructions provide a computer-readable storage media in an imaging apparatus. They include instructions that, when executed, cause the processor to implement at most an image store to store image data obtained using the imaging device. When the instructions are executed, the processor will implement an image quality controler to compare image data stored in the image storage to determine if they are suitable for use. When the instructions are executed, the processor will implement/execute the trained learning network to process image data in order to identify a clinical find in it. This will trigger an imaging apparatus notification to alert a healthcare practitioner about the clinical finding, and prompt a response with respect to a patient that is associated with the image.
Some examples show a computer-implemented way to evaluate image data taken with the mobile imaging apparatus and compare it to an image quality measurement. When the image quality measures are satisfied, the example method involves processing the image data using a learning network to identify the clinical finding in the data. An example method involves triggering an alert at the imaging device to alert a healthcare practitioner about the clinical finding. This will prompt a response with respect to the patient that is associated with the image.
The following detailed description refers to the accompanying drawings, which form a part of this disclosure. In which are shown specific examples that can be practiced. These examples provide enough detail for one skilled in art to practice the subject matter. It is also to be understood that additional examples can be used and that logical and mechanical changes as well as electrical modifications may be made to the disclosure without departing from its scope. This detailed description does not limit the scope of this disclosure. It is intended to illustrate an example implementation. You can combine features from different parts of this description to create new aspects of subject matter.
When introducing elements from various embodiments, the articles?a? ?an,? ?the,? ?said? and?the? are intended to indicate that there are one or more elements. imply that one or more elements exist. The terms ‘comprising,’? ?including,? ?Including,? These terms are meant to be inclusive. They may include additional elements.
While some examples are described in the context medical or healthcare systems, others can be used outside of the medical setting. Certain examples, such as explosive detection and non-destructive tests, can also be used in non-medical imaging.
Imaging devices (e.g. gamma camera (positron emission tomography scanner (PET), computed tomography scanner (CT) scanner), X-Ray machine (fluoroscopy machine), magnetic resonance (MR) imaging device, ultrasound scanner, etc. Generate medical images (e.g. native Digital Imaging and Communications in Medicine DICOM images) that represent the body’s parts (e.g. organs, tissues, etc.). Diagnose and/or treat disease. Volumetric data may be included in medical images, including voxels that are associated with the particular part of the body. A clinician can use medical image visualization software to annotate, measure and/or report functional and anatomical characteristics at various locations on a medical picture. A clinician might use the medical imaging software to identify areas of interest in a medical image.
Acquisition and processing of medical images data plays an important role in diagnosing and treating patients in a healthcare setting. The workflow of medical imaging can be set up, monitored and updated during the operation of medical imaging devices and workflows. Deep learning and machine technology can be used to configure, monitor and update medical imaging workflows and devices.
Some examples facilitate or provide improved imaging devices that improve coverage and diagnostic accuracy.” Some examples allow for improved image reconstruction and further processing that improves diagnostic accuracy.
Machine learning techniques, whether deep learning networks or other experiential/observational learning system, can be used to locate an object in an image, understand speech and convert speech into text, and improve the relevance of search engine results, for example. Deep learning is a subset in machine learning. It uses multiple layers of processing, including non-linear and linear transformations, to model high-level abstractions of data. Many machine learning systems start with initial features or network weights that can be changed through learning and updating the network. A deep learning network, however, trains itself to recognize?good? features. Features for analysis. Machines that use deep learning techniques to process raw data can do so much better than those using traditional machine learning techniques. Different layers of abstraction or evaluation are used to facilitate the analysis of data that has distinct themes or contains highly correlated values.
Unless the context is clear, the following terms are taken to mean the contents of the claims and specification. Deep learning is a term that refers to deep learning. Deep learning is a machine-learning technique that uses multiple data processing layers to identify different structures in data sets and classify them with high accuracy. Deep learning networks can be training networks (e.g., models or devices for training networks) that learn patterns from a variety of inputs and outputs. A deployed network, such as a network model or device for deploying deep learning networks, can also be generated from the training network and provide an output in response.
The term “supervised learning” is used to describe this type of learning. Deep learning is where the machine is given already classified data from humans. Unsupervised learning is a term that refers to unsupervised learning. This is deep learning training that does not require the machine to be given classified data, but allows the machine to detect abnormalities. What is semi-supervised learning? This is a deep-learning training method that provides the machine with a smaller amount of classified data from humans than the machine has access to more unclassified data.
Representation learning” is a term that refers to the process of learning how to represent data. This is a set of methods that transform raw data into a representation, or feature, which can be used in machine-learning tasks. Features are learned using labeled input in supervised learning.
Convolutional neural networks” is a term. “Convolutional neural networks” or?CNNs is another name. These networks of interconnected data, which are biologically inspired, are used in deep learning to detect, segment, and recognize relevant objects and regions in datasets. CNNs analyze raw data using multiple arrays and break it down into stages. Then, they examine the data for learning features.
Transfer learning” is a term that refers to the process of transferring information. A machine that stores the information needed to solve one problem correctly or incorrectly in order to solve another of the same or similar nature. Inductive learning is also known as transfer learning. Transfer learning may make use of data from past tasks, for instance.
Active learning” is a term that refers to a process of machine learning. “Active learning” is a method of machine-learning in which the machine chooses examples to receive training data from, instead of passively receiving examples selected by an external entity. As a machine learns, it can be permitted to choose examples it deems most useful for learning. This is in contrast to relying on an external expert or system to identify and provide examples.