Intel and GE Health’s, two of the biggest players in the digital healthcare sector, have in a collaborative project shown promises to significantly reduce the time between imaging and medical diagnosis. The new technology will harness the Intel Distribution of OpenVINO Toolkit which runs on Intel’s x-ray systems to help prioritizing and optimizing patient care.
Before the latest development, a longer amount of time was wasted to effectively diagnose a health condition based on the massive imaging data generated which would often lead to delayed diagnosis and treatment which in the case of critical conditions would take a toll on patient’s health. 97% of medical imaging data remains unused or unanalyzed every year according to reports. This inability of medical imaging data.
A specialized deep learning based training is undertaken where numerous labelled example images are fed to the models without specifying exact features to look for. These optimized models are able to identify details not noticeable to human eye which when integrated into GE applications with API’s from OpenVINO inference engine. The x-ray machine will acquire the images and inference engine will run the medical diagnosis. GE Healthcare’s Bigelow said that the technology’s potential will help diagnose life threatening events like pneumothorax. The technology will enable effective and efficient decision making, a good news and hope for nearly 16,000 Americans who are diagnosed every year for lung cancer. As delayed diagnosis is one of the major determining factor of cancer death rate, the technology’s promises to on time correct diagnosis will improve the patient’s chance of survival radically.
The new development is particularly promising for radiology as its trained models will be able look for target features in images such as anatomies and tumors. This will reduce retakes which will reduce patient’s unwarranted exposure to painful radiation and will make radiologists life easier by managing workload and enhanced quality of images.
Reconstructing an image from imaging modalities will become easier with deep learning based models. Future applications may include cellular microscopy, pathology, patient’s digital health records in helping develop precision based medicine and targeted drugs.