For some types of imaging, the question is not whether, but when.
Modern computers can detect patterns in clouds of data. We are all familiar with this, now that smartphones use facial recognition. Investigators in India used machine-learning techniques to develop algorithms that would analyze data from head computed tomography (CT) scans and determine whether intracranial hemorrhage, skull fracture, or mass effect were present.
Starting with roughly 300,000 CT scans, the investigators set aside 22,000 for validation of the algorithms. Using the balance of the scans, they used machine learning to derive diagnostic classification algorithms. Then they tested the accuracy of the algorithms in the set-aside scans and in two external datasets. To summarize the results, diagnostic accuracy was very good, with a…
Reviewing Author
DisclosuresConsultant/Advisory BoardPortola Pharmaceuticals, Inc.
Speaker’s BureauPeerView Institute for Medical Education
Grant/Research SupportAgency for Healthcare Research and Quality; CDC; NIH–National Center for Advancing Translational Sciences; NIH–National Institute of Allergy and Infectious Diseases (NIAID); NIH–NIAID–Antibacterial Resistance Leadership Group; Merck; Pfizer; Boehringer-Ingelheim; Shire; Portola Pharmaceuticals, Inc.; Novartis; bioMérieux; Siemens; Rapid Pathogen Screening; Magnolia; Stago; Innovative Biosensors; Molecular Detection, Inc.; Dyax Corp.; Trius Pharmaceuticals
DisclosuresConsultant/Advisory BoardPortola Pharmaceuticals, Inc.
Speaker’s BureauPeerView Institute for Medical Education
Grant/Research SupportAgency for Healthcare Research and Quality; CDC; NIH–National Center for Advancing Translational Sciences; NIH–National Institute of Allergy and Infectious Diseases (NIAID); NIH–NIAID–Antibacterial Resistance Leadership Group; Merck; Pfizer; Boehringer-Ingelheim; Shire; Portola Pharmaceuticals, Inc.; Novartis; bioMérieux; Siemens; Rapid Pathogen Screening; Magnolia; Stago; Innovative Biosensors; Molecular Detection, Inc.; Dyax Corp.; Trius Pharmaceuticals