Koji Yamashita MD, PhD, Presenter: Nothing to Disclose

Akio Hiwatashi MD, Abstract Co-Author: Nothing to Disclose

Osamu Togao MD, PhD, Abstract Co-Author: Nothing to Disclose

Kazufumi Kikuchi MD, Abstract Co-Author: Nothing to Disclose

Masatoshi Kondo, Abstract Co-Author: Nothing to Disclose

Hiroshi Sugimori, Abstract Co-Author: Nothing to Disclose

Takashi Yoshiura MD, PhD, Abstract Co-Author: Nothing to Disclose

Hiroshi Honda MD, Abstract Co-Author: Nothing to Disclose

It is important to predict neurological outcome in patients with non-traumatic coma. our purpose was to evaluate the prognostic utility of CT histogram analysis with an automated whole-brain extraction algorithm in comatose patients.

We retrospectively studied 138 consecutive comatose patients who were admitted to our intensive care unit and underwent brain CT. The patients were classified into good (n = 43; M:F = 18:25) and poor (n = 95; M:F = 48:47) outcome groups. All CT images were obtained using a 64-detector-row CT scanner with a slice thickness of 4.0 mm. From the whole-brain CT images, a brain region was extracted using our original automated algorithm for the subsequnt histogram analysis. The obtained histogram statistics (mean CT value, kurtosis and skewness) as well as clinical parameters were compared between the good and poor outcome groups using the Mann-Whitney U test. In addition, ROC analysis was performed for the discrimination between the 2 goups for each parameter.

The mean CT value was significantly higher in the good outcome group (mean±SD = 34.6±1.47 HU) than in the poor outcome group (mean±SD = 33.9±1.97 HU) (p<0.05). In addition, the kurtosis and age were significantly lower in the good outcome group (mean kurtosis±SD = -0.49±0.12, mean age±SD = 54.1±21.4 years) than in the poor outcome group (mean kurtosis±SD = -0.34±0.21, mean age±SD = 63.7±18.6 years) (p<0.001 and p<0.05, respectively). The AUC values for the kurtosis, mean CT value, and age were 0.717, 0.608, and 0.625, respectively. A combination of the 3 parameters increased the diagnostic performance (AUC = 0.799).

Histogram analysis of whole-brain CT images with our automated extraction algorithm is useful for assessing the prognosis of comatose patients.

Histogram analysis method tend to be more reproducible compared with manual region-of-interest placement. Our study revealed that histogram parameters as well as age can help predict the neurological outcome of comatose patients.

Yamashita, K, Hiwatashi, A, Togao, O, Kikuchi, K, Kondo, M, Sugimori, H, Yoshiura, T, Honda, H, Prognostic Value of CT Histogram Analysis in Comatose Patients: Evaluation Using Automated Whole-brain Extraction Algorithm.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14007762.html