Machine Learning in Radiology: Applications Beyond Image Interpretation.

Much attention has been given to machine learning and its perceived impact in radiology, particularly in light of recent success with image classification in international competitions. However, machine learning is likely to impact radiology outside of image interpretation long before a fully functional "machine radiologist" is implemented in practice. Here, we describe an overview of machine learning, its application to radiology and other domains, and many cases of use that do not involve image interpretation. We hope that better understanding of these potential applications will help radiology practices prepare for the future and realize performance improvement and efficiency gains.

Abnormal connectivity in the posterior cingulate and hippocampus in early Alzheimer's disease and mild cognitive impairment.

BACKGROUND: Brain imaging studies of early Alzheimer's disease (AD) have shown decreased metabolism predominantly in the posterior cingulate cortex (PCC), medial temporal lobe, and inferior parietal lobe. This study investigated functional connectivity between these regions, as well as connectivity between these regions and the whole brain. METHODS: Functional magnetic resonance imaging (fMRI) and diffusion tensor imaging (DTI) studies were performed in subjects with early AD, mild cognitive impairment (MCI), and normal controls. RESULTS: The data indicate both decreased fiber connections and disrupted connectivity between the hippocampus and PCC in early AD. The MCI group showed reduced fiber numbers derived from PCC and hippocampus to the whole brain. CONCLUSIONS: The fMRI and DTI results confirmed decreased connectivity from both the PCC and hippocampus to the whole brain in MCI and AD and reduction in connectivity between these two regions, which plausibly represents an early imaging biomarker for AD.

The Self Test: a screening tool for dementia requiring minimal supervision.

BACKGROUND: The main purpose of this study was to correlate a newly developed, simple, but comprehensive screening test for dementia called the Self Test (ST) with the Mini-mental State Examination (MMSE), and establish the reliability and validity of the ST. METHODS: The ST was administered to 42 consecutive new referrals to the memory clinic at the Cole Neuroscience Center, University of Tennessee Medical Center (UTMC), Knoxville, TN, U.S.A. and 41 age-matched control subjects at regular intervals over a 12-month period. RESULTS: The total ST scores correlated well with cognitive severity as measured by the MMSE (r = 0.71, p < 0.01). The area under the receiver-operating characteristic curve (AUROCC) in the group of patients with Alzheimer's disease (AD) with less than a high school education was 0.80, and 0.89 in those with at least a high school education. Reliability analysis showed a mean inter-item correlation of 0.31 for patients with AD, and 0.47 for normal participants. Cronbach's alpha was calculated to be 0.70 for the AD group. Test-retest reliability was determined using intraclass correlation coefficients (ICC = 0.93, p < 0.01). CONCLUSIONS: The ST is an internally consistent, reliable and valid screening test for cognitive impairment in persons exhibiting early symptoms of dementia. In addition, the ST requires minimal nonprofessional supervision and may be administered by an untrained person.

Pelvic imaging in the stable trauma patient: is the AP pelvic radiograph necessary when abdominopelvic CT shows no acute injury?

The purpose of the study was to determine the utility of anteroposterior (AP) pelvic radiographs in stable trauma patients who will undergo or have undergone abdominopelvic CT as part of the initial trauma imaging evaluation. Radiology reports of all stable trauma patients who underwent both abdominopelvic CT and AP pelvic radiograph from 25 January through 30 April, 2003 were reviewed for findings of acute pelvic injuries. A total of 509 consecutive patients were included in this series. Of these, 449 patients (88.2%) had no acute pelvic injury revealed by abdominopelvic CT. CT showed 163 acute injuries in 60 patients. AP radiographs showed 132 acute injuries in 52 patients. No patients with a negative CT had an acute finding on the radiograph. There were eight false-negative pelvic radiographs (negative predictive value 98.25%). CT is highly accurate in excluding acute osseous pelvic injuries. In the stable trauma patient whose CT does not reveal an acute pelvic injury or who is scheduled to undergo an abdominopelvic CT as part of the initial imaging evaluation, the pelvic radiograph may be unnecessary.