Overview

Detection of acute intracranial hemorrhage (AIH) is a primary task in image interpretation of computer tomography (CT) of brain for patients suffering from acute neurological disturbance or head injury. Although CT readily depicts AIH, interpretation can be difficult especially when the lesion is inconspicuous or the reader is inexperienced. We have developed a computer aided detection (CAD) system that improves diagnostic accuracy of small AIH on brain CT.

Methods

Intracranial contents are first segmented by thresholding, morphological operations, and continuity analysis. AIH candidates are then segmented based on top-hat transformation, left-right asymmetry, and adjustment for CT cupping artifacts Then true AIH is differentiated from mimicking normal variants or artifacts by a knowledge based classification system incorporating rules that make use of quantified imaging features and anatomical information. (Figure 1).

Figure 1. Schematic diagram of the AIH CAD system. Intermediary outputs of an image showing right basal ganglia hemorrhage illustrate the effect of individual steps. Identified AIH is showed in RED.

Results

A total of 186 clinical cases, including 62 CT studies showing small (< 1cm) AIH, and 124 controls, were retrospectively collected. 40 positive cases and 80 controls were used for the training of the CAD. Twenty-two positive cases and 44 controls were used in the validation of the CAD system. Regions of AIH identified by two experienced radiologists were used as gold standard. The size of individual AIH volume was also recorded. Results demonstrate that per patient basis, sensitivity and specificity for both the training and validation cases can achieve over 90% and 85% respectively.

Clinical Relevance

A multiple-reader multiple-case receiver operating characteristic (ROC) study was performed. Twenty clinicians, including 7 emergency room (ER) physicians 7 radiology residents and 6 specialists, were recruited as readers of 60 sets of brain CT, including 30 cases that show AIH smaller than 1cm, and 30 controls. Significantly improved performance is observed in ER physicians, average area under the ROC curve (A) increased from 0.83 to 0.95 when they make the diagnosis without and with the support of CAD. And for radiology residents, and specialists, A increased from 0.95 to 0.98, and from 0.97 to 0.98, respectively.
We are in the process of implementing this CAD method for daily clinical operation to aid physicians to improve their AIH diagnostic performance.

Reference

Tao Chan and H.K. Huang, “ Computer Assisted Diagnosis of Small Acute Intracranial Hemorrhage using CT”, RSNA InfoRAD Exhibit, RSNA, 2006.