Cardiology
November 1, 2021AVA Bracelet
November 1, 2021Intel®
Intel® designs allow its users to deploy applications both quickly and cost-effectively. Users will notice experiences and solutions to manage each need...
Today, implementing artificial intelligence (AI) is important to serve customers better and unlock new business worth. Deploying the infrastructure required to support AI at scale remains a challenge for several enterprises, given the complexness of achieving high performance and accuracy while managing prices and utilization. With Intel’s design, users can deploy applications into production both quickly and cost-effectively. With access to Intel’s world partner ecosystem, everyone will notice experiences and solutions to manage each AI need.
The Four Paths to AI Success
Perceptions of AI within the enterprise are evolving. Every organization is different in terms of AI readiness.
The character of people’s desires, the maturity of any existing data, science capabilities, and the level of expertise available in-house are just some of the factors that impact AI strategies.
In this article, we will explore some strategies Intel is applying.
For companies new to AI that need fast time-to-market: Intel collaborates with system integrators, software, or hardware providers to implement ready-made solutions.
For companies who run a cloud-first strategy and have elastic needs: Intel uses cloud-based AI instances provided by leading cloud service providers.
For companies with mature data science skills in-house or need fast time-to-market: Intel implements on-premises solutions using pre-designed and validated options.
For companies who have mature data science skills in-house or need a finely tailored solution to meet specific needs: Intel builds tailored AI solutions in-house for each client. (1)
Intel Leader of IT Solutions
IT at Intel plays a central role in increasing the value of Intel’s business. This includes integrating and expanding AI and machine learning to critical business operations, capitalizing on the flexibility of Intel’s hybrid cloud, and reducing technical debt.
Health IT and Analytics: From medical history to genomics, clinicians and payers will gain fresh insight into a wide range of applications, from better tailored individual treatment to statistical modeling for large population cohorts, resulting in increased patient outcomes and better provider and payer cost use. Healthcare technology application built on Intel includes:
Medical imaging: for CT, MRI, PET, X-ray, ultrasound, and endoscopy applications, Intel offers a variety of computational options.
Analytics: Intel’s big new data systems, AI models, edge computing, and other analytics in health information technology are built on Intel’s base.
Lab and Life Sciences: Laboratory breakthroughs lead to new strategies in healthcare and medicine, including customized treatments based on a patient’s genetic data. Laboratory automation in hospitals and health systems enables high accuracy and fast turnaround time for diagnostic testing. It entails a range of technologies designed to automate high-volume and manual processes in clinical or other laboratories. These technologies involve lab robotics and artificial intelligence, including machine learning, deep learning, and computer vision.
Telemedicine: For remote care, including vitals monitoring, chronic disease management, and specialist consultations.
Robotics: New categories of robots powered by Intel technologies. Surgeons may use AI to conduct noninvasive procedures to understand their patients’ internal organs better. Other strategies include the aid of manual operations by disinfection and logistics robotics. (2)
Intel as a Business
Intel reported in its Second-Quarter 2020 Financial Results a revenue of $19.7 billion, 20% up over the previous year. Intel achieved record second-quarter revenue with 34% data-centric revenue growth and 7% P-centric revenue growth year after year.
Powerful sales of cloud, laptop, memory, and 5G devices drove these results to an environment where digital services and computing efficiency were critical. Intel’s memory company set a new sales milestone in the period, and Intel’s 5G network connectivity portfolio gained traction with customers. (3)
Intel’s distribution of Open VINOTM Toolkit Optimizes Deep Learning Performance and Healthcare Imaging
The application of Artificial Intelligence to medical imaging can improve the process on multiple levels, like productivity and enhanced imaging quality. Deep learning is a form of AI that aims to provide the highest quality of medical imaging.
The Intel distribution of Open VINO toolkit comprises Intel processors and accelerators that rapidly provide deep learning models.
Despite different hardware platforms that can train models, the toolkit imports trained models from different deep learning systems such as Caffe, MXNet, TensorFlow, and ONNX.
GE Healthcare delivers medical imaging equipment and other healthcare technologies. It has collaborated with Intel to test the overall performance of GE’s deep learning solutions, which group-scanned image slices. Thus, relevant images are easy to find, and their use for research or clinical comparison would become much more practical. (4)
Intel as a Business
Segmentation of the brain using MRI is important for diagnosing brain tumors; this has been, traditionally, a manual process. The development of an automatic method could save time, decrease errors, and improve accessibility. From Princeton University and Intel Labs, a Deep Learning system was developed focusing on the structural properties of the brain. More precisely, the principle that the left and right hemispheres of a healthy brain are symmetrical and the presence of any high-level asymmetry implies abnormality.
Researchers trained the system with multimodal MRIS scans of 285 subjects with brain tumors, randomly partition 210 of those images into two parts, with 80% for training and 20% for validation, and compared the results with U-net, a standard AI system.
The results showed that the new model takes fewer epochs to converge and showed a higher Dice score. This suggests that encoding symmetry in the neural network helps extract characteristics relevant to brain tumor segmentation. With the potential of adding the system to the standard U-net to improve performance. (5)
The Use of Artificial Intelligence in the Development of a Computer-Assisted Surgery (CAS) System
Implementing artificial intelligence systems for the reconstruction, structure rendering, and surgery planning, helps doctors optimize surgery plans, improve tumor resection rates, and safety, therefore benefiting patients. A precise 3D reconstruction of the lungs based on anatomic segmentation using AI can analyze and identify structures, tissues, lesions, blood vessels, and bronchial tubes, achieving an accurate visualization. Precision segmentation allows lesion removal at minimal cost with improved long-term survival and better quality of life.
A group of researchers from Hisense Medical developed a CAS system that can provide a complete lung image reconstruction with precision and efficacy. The software can process and segment CT and MR images and provide surgical simulation tools for cutting, mapping, and blood vessel analysis.
The system is trained using standard public database and checkpoint files for inference result validation and performance evaluation.
Hisense Medical CAS System provides a 360-degree, quantifiable, observable, and interactive image model. Doctors can make close-up views, rotate the model, adjust transparency, and zoom in and out. The CAS system has great potential in the management of pulmonary diseases assisting doctors to improve the quality of clinical surgery. (6)
Intel Supports COVID-19 Research
Intel’s architecture-optimized genomics analytics framework aims that by partnering with BIH, researchers will find solutions that will help overcome the COVID-19 pandemic. To achieve its goal of better understanding how SARS- CoV-2 works, the BIH – Center for Digital Health set out to examine samples from the lower respiratory tract of 52 samples originating from the upper and lower respiratory tract of 40 individuals—resulting in nearly 220,000 single cells. This study has revealed the type of cells that have the receptors the virus infects and how the host cells respond to the subsequent virus infection. (7)
