IMedis

Recent years have seen a real revolution in the world of medicine. The introduction of artificial intelligence (AI) technologies enables the integration and analysis of data from various sources, with the aim of optimizing diagnostic and preventive medicine.

The medical world is making huge strides in developing innovative and effective treatments for life-threatening diseases, but in many cases, the most influential factor in the chances of recovery is not necessarily the quality of the treatment, but the stage of diagnosis. As the amount of medical information accumulated in each medical examination increases, the major challenge now facing the medical system is strengthening and completing physicians' diagnostic capabilities through advanced technologies.

The revolutionary platform of IMedis uses advanced AI algorithms to accurately and quickly analyze medical imaging together with text from the medical report, and to ensure that important medical information does not go unnoticed.

Need

In medicine, despite the astonishing rate of deaths due to diagnostic errors, there are no standard, automatic safeguards unlike every other domain in our life. It is specifically relevant to radiology which suffers from a steady rate of errors in diagnosis for decades. Survival after a cancer diagnosis can be 40% higher after early as opposed to late diagnosis. An NIH study found that up to 85% of lung cancer patients are diagnosed at a very late stage, at which point the illness is rarely curable.

Advanced medical imaging tools, most notably CT scans, have become the primary diagnostic tools in recent years. These tests have tremendous potential for early disease diagnosis, but studies show that over 6% of CT scans have an important finding, which requires medical attention, but is missed by the radiologist. In the US alone, there are approximately 10,000 "missed" tests every day.

Since these medical conditions are usually without obvious symptoms in their early stages, most of these diagnoses are incidental and discovered when imaging tests are performed for other medical reasons. For example, a CT scan of the chest performed after a car accident may incidentally detect an existing liver lesion that requires further and important in-depth examination.

Solution

The founding team of IMedis have a deep familiarity with the existing work processes in imaging departments. They recognized the importance of radiologists being able to identify important findings that are often not the focus of the examination and, therefore, less likely to be diagnosed. At the same time, in order to leverage such a platform in the clinical setting, it must be quickly and easily implemented, must not create a high workload for the medical team and must significantly improve the output and quality of the radiologist’s reports.

To that end, IMedis has developed Duali-Q, an AI-based software system that automatically analyzes, in parallel to the radiologist's work, all CT examinations at the hospital, in order to identify suspicious findings that require medical attention. In addition, the system simultaneously analyzes diagnostic reports written by radiologists in real time. When the system detects an important finding that appears in a scan, but does not appear in the report, it alerts the radiologist in real time, allowing him/her to quickly and efficiently re-evaluate the images and include the additional findings in the medical report.

The philosophy behind Duali-Q is the understanding that the place of technology is not to replace physicians but to augment their capabilities, since in most cases, radiologists perform accurate and quick diagnosis. To that end, IMedis offers a parallel diagnostic tool to the one currently available. In this way, the system decides to intervene in real time, alerting only if a key finding is missed, and thus offering radiologists a tool that enhances their capabilities and quality of output.

Technology

Deep learning is a subfield of artificial intelligence that has evolved in recent years and made a significant breakthrough in a variety of technological areas, most notably in automatic processing and the analysis of images and text. Deep learning algorithms are the technological foundation of IMedis. The company has developed a unique platform, based on deep learning algorithms, for processing CT scans and medical text from radiological reports.

Using a series of deep learning algorithms of Convolutional Neural Networks (CNN), the system can automatically detect, mark, classify and measure suspected lesions appearing in CT scans, including those smaller than 1 cm, which presents a challenge even for trained and experienced physicians.

In addition, IMedis has developed Recurrent Neural Networks (RNN) deep learning algorithms for text analysis, with which the system is able to identify any clinical findings in the radiological report and classify them as positive references (for example: "a suspicious liver lesion of 8 mm diameter") or negative references (for example: "No suspicious lesions were found in the liver").

To protect the innovation, IMedis has several filed patents, with emphasizes on the unique combination described above.