Computer-Aided Diagnosis Scheme for Detection of Unruptured Intracranial Aneurysms in MR Angiography

Computer-aided diagnosis improves detection of small intracranial aneurysms on MRA in a clinical setting

BACKGROUND AND PURPOSE

MRA is widely accepted as a noninvasive diagnostic tool for the detection of intracranial aneurysms, but detection is still a challenging task with rather low detection rates. Our aim was to examine the performance of a computer-aided diagnosis algorithm for detecting intracranial aneurysms on MRA in a clinical setting.

MATERIALS AND METHODS

Aneurysm detectability was evaluated retrospectively in 48 subjects with and without computer-aided diagnosis by 6 readers using a clinical 3D viewing system. Aneurysms ranged from 1.1 to 6.0 mm (mean = 3.12 mm, median = 2.50 mm). We conducted a multireader, multicase, double-crossover design, free-response, observer-performance study on sets of images from different MRA scanners by using DSA as the reference standard. Jackknife alternative free-response operating characteristic curve analysis with the figure of merit was used.

RESULTS

For all readers combined, the mean figure of merit improved from 0.655 to 0.759, indicating a change in the figure of merit attributable to computer-aided diagnosis of 0.10 (95% CI, 0.03-0.18), which was statistically significant (F(1,47) = 7.00, P = .011). Five of the 6 radiologists had improved performance with computer-aided diagnosis, primarily due to increased sensitivity.

CONCLUSIONS

In conditions similar to clinical practice, using computer-aided diagnosis significantly improved radiologists' detection of intracranial DSA-confirmed aneurysms of ≤6 mm.

Computer-aided diagnosis for detection of lacunar infarcts on MR images: ROC analysis of radiologists' performance

The purpose of this study was to retrospectively evaluate radiologist performance in detection of lacunar infarcts on T1- and T2-weighted images, without and with the use of a computer-aided diagnosis (CAD) scheme. Thirty T1-weighted and 30 T2-weighted MR images obtained from 30 patients were used for assessing observer performance. These images were acquired using the fast spin-echo sequence with a 1.5-T MR imaging scanner. The group included 15 patients (age range, 48-83 years; mean age, 67.2 years; 10 men and five women) with a lacunar infarct and 15 patients (age range, 39-76 years; mean age, 64.0 years; eight men and seven women) without lacunar infarcts. Nine radiologists participated in the study. The radiologists initially interpreted the T1- and T2-weighted images without and then with the use of CAD, which indicated their confidence levels regarding the presence (or absence) of lacunar infarcts and the most likely position of a lesion on each MR scan. The observers' performance without and with the computer output was evaluated by performing receiver operating characteristic analysis. For the nine radiologists, the mean area under the best-fit binormal receiver operating characteristic curve plotted for unit square values of radiologists who interpreted the images without and with the scheme were 0.891 and 0.937, respectively. The performance of the radiologists improved significantly when they used the computer output (p=0.032). The CAD scheme has potential to improve the accuracy of radiologists' performance in detection of lacunar infarcts.

Three-dimensional morphological analysis of intracranial aneurysms: a fully automated method for aneurysm sac isolation and quantification

PURPOSE

Morphological descriptors are practical and essential biomarkers for diagnosis and treatment selection for intracranial aneurysm management according to the current guidelines in use. Nevertheless, relatively little work has been dedicated to improve the three-dimensional quantification of aneurysmal morphology, to automate the analysis, and hence to reduce the inherent intra and interobserver variability of manual analysis. In this paper we propose a methodology for the automated isolation and morphological quantification of saccular intracranial aneurysms based on a 3D representation of the vascular anatomy.

METHOD

This methodology is based on the analysis of the vasculature skeleton's topology and the subsequent application of concepts from deformable cylinders. These are expanded inside the parent vessel to identify different regions and discriminate the aneurysm sac from the parent vessel wall. The method renders as output the surface representation of the isolated aneurysm sac, which can then be quantified automatically. The proposed method provides the means for identifying the aneurysm neck in a deterministic way. The results obtained by the method were assessed in two ways: they were compared to manual measurements obtained by three independent clinicians as normally done during diagnosis and to automated measurements from manually isolated aneurysms by three independent operators, nonclinicians, experts in vascular image analysis. All the measurements were obtained using in-house tools. The results were qualitatively and quantitatively compared for a set of the saccular intracranial aneurysms (n = 26).

RESULTS

Measurements performed on a synthetic phantom showed that the automated measurements obtained from manually isolated aneurysms where the most accurate. The differences between the measurements obtained by the clinicians and the manually isolated sacs were statistically significant (neck width: p <0.001, sac height: p = 0.002). When comparing clinicians' measurements to automatically isolated sacs, only the differences for the neck width were significant (neck width: p <0.001, sac height: p = 0.95). However, the correlation and agreement between the measurements obtained from manually and automatically isolated aneurysms for the neck width: p = 0.43 and sac height: p = 0.95 where found.

CONCLUSIONS

The proposed method allows the automated isolation of intracranial aneurysms, eliminating the interobserver variability. In average, the computational cost of the automated method (2 min 36 s) was similar to the time required by a manual operator (measurement by clinicians: 2 min 51 s, manual isolation: 2 min 21 s) but eliminating human interaction. The automated measurements are irrespective of the viewing angle, eliminating any bias or difference between the observer criteria. Finally, the qualitative assessment of the results showed acceptable agreement between manually and automatically isolated aneurysms.

Computer-aided diagnosis scheme for classification of lacunar infarcts and enlarged Virchow-Robin spaces in brain MR images

The detection of asymptomatic lacunar infarcts on magnetic resonance (MR) images is important because their presence indicates an increased risk of severe cerebral infarction. However, accurate identification of lacunar infarcts on MR images is often hard for radiologists because of the difficulty in distinguishing lacunar infarcts and enlarged Virchow-Robin spaces. Therefore, we developed a computer-aided diagnosis (CAD) scheme for the classification of lacunar infarcts and enlarged Virchow-Robin spaces. Our database consisted of T1- and T2- weighted images obtained from 109 patients. The locations of lacunar infarcts and enlarged Virchow-Robin spaces were determined by an experienced neuroradiologist. It included 89 lacunar infarcts and 20 enlarged Virchow-Robin spaces. We first enhanced the lesions in T2-weighted image by using the white top-hat transformation. A gray-level thresholding was then applied to the enhanced image for the segmentation of lesions. From the segmented lesions, we determined image features, such as size, shape, location, and signal intensities in T1- and T2- weighted images. A neural network was then employed for distinguishing between lacunar infarcts and enlarged Virchow-Robin spaces. Our computerized method was evaluated by using a leave-one-out method. The result indicated that the area under the ROC curve was 0.945. Therefore, our CAD scheme would be useful in assisting radiologists for diagnosis of silent cerebral infarctions in MR images.

Improvement of automated detection method of lacunar infarcts in brain MR images

The detection of asymptomatic lacunar infarcts on magnetic resonance (MR) images are important tasks for radiologists to ensure the prevention of sever cerebral infarction. However, their accurate identification is often difficult task. Therefore, the purpose of this study is to develop a computer-aided diagnosis scheme for the detection of lacunar infarcts. Our database consisted of 1,143 T1- and 1,143 T2-weighted images obtained from 132 patients. We first segmented the cerebral region in the T1- weighted image by using a region growing technique. For identifying the initial lacunar infarcts candidates, white top-hat transform and multiple-phase binarization were then applied to the T2- weighted image. For eliminating false positives (FPs), we determined 12 features, i.e., the locations x and y, density differences in the T1- and T2- weighted images, nodular components (NC), and nodular & linear components (NLC) from a scale 1 to 4. The NCs and NLCs were obtained using filter bank technique. The rule-based scheme and a neural network with 12 features were employed as the first step for eliminating FPs. The modular classifier was then used for eliminating three typical sources of FPs. As a result, the sensitivity of the detection of lacunar infarcts was 96.8% with 0.30 FP per image. Our computerized scheme would assist radiologists in identifying lacunar infarcts on MR images.

Automated classification of cerebral arteries in MRA images and its application to maximum intensity projection

Detection of unruptured aneurysms is a major task in magnetic resonance angiography (MRA). However, it is difficult for radiologists to detect small aneurysms on the maximum intensity projection (MIP) images because adjacent vessels may overlap with the aneurysms. Therefore, we proposed a method for making a new MIP image, the SelMIP image, with the interested vessels only, as opposed to all vessels, by manually selecting a cerebral artery from a list of cerebral arteries recognized automatically. By using our new SelMIP viewing technique, the selected vessel regions can also be observed from various directions and would further facilitate the radiologists in detecting small aneurysms. For automated classification of cerebral arteries, two 3D images, a target image and a reference image, are compared. Image registration is performed using the global matching and feature correspondence techniques. Segmentation of vessels in the target image is performed using the thresholding and region growing techniques. The segmented vessel regions were classified into eight cerebral arteries by calculating the Euclidean distance between a voxel in the target image and each of the voxels in the labeled eight vessel regions in the reference image. In applying the automated cerebral arteries recognization algorithm to thirteen MRA studies, results of 10 MRA studies were evaluated as clinically acceptable. Our new viewing technique would be useful in assisting radiologists for detection of aneurysms and for reducing the interpretation time.