Modernday medical imaging exams have become a critical diagnostic tool for conditionsof all kinds – from detecting the earliest breast cancers, long before a tumorcould grow large enough for a woman to feel a lump in her own body, tofinding malformations in the hearts of tiny babies monthsbefore they’re ready to be born. The instruments developed to look inside thebody to capture these images become more powerful by the day. “A patient canwalk in and in just a few minutes, generate a gigabyte worth of data,” says James Gee, PhD, an associate professor ofRadiologic Science and Computer and Information Science, who directsthe Penn Image Computing and Science Laboratory(PICSL).
Butthe increasingly detailed pictures typically still require a human being – aradiologist, often with specialized additional training in areas likeneuroradiology, musculoskeletal or cardiothoracic imaging – to examine theimages, tease out the answers they hold, and use the information to arrive at adiagnosis, which will ultimately be used to shape the treatment plan.
Thisprocess can be painstaking: Mapping out the locations of all the differentparts of a patient’s brain captured via magnetic resonance imaging, forinstance, might take even an experienced clinician nearly a whole day tocomplete. Using a new computer algorithm developed by medical imagingresearchers at Penn Medicine, however, that labeling process happensautomatically, taking “zero time.” And perhaps more importantly, the computer’sanswers are extraordinarily accurate, Gee says. If heshowed unmarked images of a brain that was labeled manually versus one that wassegmented automatically by the new algorithm – as shown in the image above --and presented them to experts, they would “hard-pressed to pick which was doneby the human.”
Thenew algorithm, which automatically finds and labels anatomical strictures inMRI scans, won first place in a Grand Challenge competition held during therecent International Conference on Medical ImageComputing and Computer Assisted Intervention. Authors of the submission includePaul Yushkevich, PhD, Hongzhi Wang, PhD, and Brian Avants, PhD. In addition tothe team’s win, more than a third of the entries from teams across the worldwere built using open-source image registration software developed in thePICSL, which won first place in a previous MICAAI competition.
Thetechnology the algorithm is based on – homegrown at Penn beginning in the late 1980s [a1] --is already widely in use at Penn Medicine in clinical research. Trialscomparing changes in the brains of patients with Alzheimer’s disease to thoseof normal control subjects are one example. In the future, Gee said he hopesthe time saved on manual labeling of radiological images of all kinds – acrossall types of imaging technologies, from those generated during obstetricalultrasounds to cardiac CT scans – will lead to expedited diagnoses and quickertreatment for patients.
Intime, he envisions that actionable diagnostic information will be obtained muchquicker, making some of the delays associated making diagnoses a thing of thepast. “Imagine taking someone’s blood pressure or temperature and it being somecomplicated process that took a long time, instead of getting just a numberinstantly,” Gee says. “Technology like ours is the wave of the future inmedical imaging.”