University electrical engineers have patented a novel search engine that could have major implications for radiology.The researchers claim the method can search image archives for certain anatomic or pathological properties, such as specific types of
University electrical engineers have patented a novel search engine that could have major implications for radiology.
The researchers claim the method can search image archives for certain anatomic or pathological properties, such as specific types of fractures, skeletal areas, or vascular features.
"This technique is not restricted to flat text or genome data," said Ronald Indeck, Ph.D., director of the Center for Security Technologies in the electrical engineering department at Washington University in St. Louis. "The data can be radiological images."
While the system is not yet ready to search image databases for, say, chest studies with solitary pulmonary nodules of a certain size or shape, this is not beyond its scope.
"That is exactly where we will be soon," Indeck said.
The system will include a model for the background that is then used to compare and detect anomalies. Currently, it has the capability of searching unstructured data in the form of flat text (genetic sequences, textual files, etc.) - 200 times faster than other popular search engines.
"Today, images are tagged with metadata, which is extremely limited," Indeck said. "It might pertain to pathology, or the person's name, but once you're through the metadata you're done. You can't search anymore."
His system works differently. Using what he calls "reconfigurable hardware," Indeck's engine makes use of existing computing components and puts them to work in novel ways.
"Having mass data storage so affordable has enabled enormous amounts of data to be archived," Indeck said. "However, these databases far exceed the amount of memory available to a processor so that searching becomes a serious challenge."
Conventional engines build reverse indexes, like those in books.
"The problem is that much stored data - especially medical data - are not indexable. Even if you could build an index, it would require so much memory to process, it would be a losing proposition," Indeck said.
Instead of translating the magnetic signal into bits that are then indexed by the computer processor, Indeck recruits the high-speed parallel magnetic sensing systems already present in modern magnetic storage devices to facilitate searches.
He searches these databases directly, without processor, memory, or bandwidth limitations. Data are searched in place and not moved off the drive into memory. The results improve the speed and cost of performing approximate matches within large spaces by orders of magnitude.
According to Indeck, the system so far has performed well with medical databases such as the American College of Radiology Imaging Network, a National Cancer Institute-sponsored cooperative group established to perform multi-institutional clinical trials in diagnostic imaging related to cancer.
MRI-Based AI Radiomics Model Offers 'Robust' Prediction of Perineural Invasion in Prostate Cancer
July 26th 2024A model that combines MRI-based deep learning radiomics and clinical factors demonstrated an 84.8 percent ROC AUC and a 92.6 percent precision-recall AUC for predicting perineural invasion in prostate cancer cases.
Breast MRI Study Examines Common Factors with False Negatives and False Positives
July 24th 2024The absence of ipsilateral breast hypervascularity is three times more likely to be associated with false-negative findings on breast MRI and non-mass enhancement lesions have a 4.5-fold likelihood of being linked to false-positive results, according to new research.
Can Polyenergetic Reconstruction Help Resolve Streak Artifacts in Photon Counting CT?
July 22nd 2024New research looking at photon-counting computed tomography (PCCT) demonstrated significantly reduced variation and tracheal air density attenuation with polyenergetic reconstruction in contrast to monoenergetic reconstruction on chest CT.