Computer-Aided Diagnosis in Medical Imaging: Historical Review, Current Status and Future Potential

The incidence of subarachnoid haemorrhage (SAH) is 6-8 per 100 000 person years, peaking in the sixth decade. SAH, mostly due to rupture of an intracranial aneurysm, accounts for a quarter of cerebrovascular deaths. Aneurysms increase in frequency with age beyond the third decade, are 1.6 times more common in women and are associated with a number of genetic conditions. Prospective autopsy and angiographic studies indicate that between 3.6 and 6% of the population harbour an intracranial aneurysm. Studies have found an increased rate of SAH in first degree relatives of SAH patients (relative risk 3.7-6.6). In affected families, the most frequent relationship between sufferers is sibling to sibling. The rupture rate of asymptomatic aneurysms was thought to be 1-2% per annum, but the recent International Study of Unruptured Intracranial Aneurysms found that the rupture rate of small aneurysms was only 0.05% per annum in patients with no prior SAH, and 0.5% per annum for large (>10 mm diameter) aneurysms and for all aneurysms in patients with previous SAH. Non-invasive tests such as magnetic resonance angiography (MRA), computed tomographic angiography (CTA) and transcranial Doppler (TCD) have been advocated as alternatives to intra-arterial digital subtraction angiography to screen for aneurysms. Although all are promising techniques, the quality of data testing their accuracy is limited. Overall reported sensitivity for CTA and MRA (TCD is poorer) was 76-98% and specificity was 85-100%, but many subjects had an aneurysm or recent SAH, which could overestimate accuracy. CTA and MRA are much poorer methods for the detection of aneurysms <5 mm diameter, which account for up to one-third of unruptured aneurysms. Elective surgical clipping of asymptomatic aneurysms has a morbidity of 10.9% and mortality of 3. 8%. Treatment of aneurysms by Guglielmi coils, for which there is less long-term follow-up available, has a 4% morbidity and 1% mortality, but only achieves complete aneurysm occlusion in 52-78% of cases. There has been interest in screening for aneurysms, but the indication for, and cost effectiveness of screening are unclear because aneurysm prevalence varies, rupture rate is low, non-invasive imaging tests are not yet accurate enough to exclude small aneurysms and the morbidity and mortality for elective surgical treatment of unruptured aneurysms is high. There may be a limited role for investigation of high risk subgroups. Ideally, screening in such subgroups should be tested in a randomized trial. The avoidance of risk factors for aneurysms such as smoking, hypertension and hypercholesterolaemia should be part of the management of at-risk subjects.

The detection and management of unruptured intracranial aneurysms

Background and Purpose The purpose of this study was to assess the influence of race, sex, and other risk factors on the location of atherosclerotic occlusive lesions in cerebral vessels. Previous angiographic studies of patients with stroke or transient ischemic attack (TIA) suggest that extracranial atherosclerosis is more common in whites and intracranial disease is more common in blacks. Noninvasive techniques such as duplex ultrasound, transcranial Doppler (TCD), and magnetic resonance angiography (MRA) allow vascular assessment of a more representative proportion of patients than does conventional angiography alone. Methods Consecutive patients evaluated at a community hospital for stroke or TIA over a 2-year period were reviewed. Lesions were defined as a 50% or greater atherosclerotic stenosis by angiography, duplex ultrasound, or TCD, or a moderate stenosis by MRA. Results Whites were more likely than blacks to have extracranial carotid artery lesions (33% versus 15%, P=.001), but the proportion ...

Race and Sex Differences in the Distribution of Cerebral Atherosclerosis

BACKGROUND AND PURPOSE: Although digital subtraction angiography (DSA) provides excellent visualization of the intracranial vasculature, it has several limitations. Our purpose was to evaluate the ability of helical CT angiography (CTA) to help detect and quantify intracranial stenosis and occlusion compared with DSA and MR angiography (MRA). METHODS: Twenty-eight patients underwent CTA, DSA, and 3D time-of-flight (TOF) MRA for suspected cerebrovascular lesions. All three studies were performed within a 30-day period. Two readers blinded to prior estimated or calculated stenoses, patient history and clinical information examined 672 vessel segments. Lesions were categorized as normal (0–9%), mild (10–29%), moderate (30–69%), severe (70–99%), or occluded (no flow detected). DSA was the reference standard. Unblinded consensus readings were obtained for all discrepancies. RESULTS: A total of 115 diseased vessel segments were identified. After consensus interpretation, CTA revealed higher sensitivity than that of MRA for intracranial stenosis (98% versus 70%, P CONCLUSION: CTA has a higher sensitivity and positive predictive value than MRA and is recommended over TOF MRA for detection of intracranial stenosis and occlusion. CTA has a high interoperator reliability. CTA is superior to DSA in the evaluation of posterior circulation steno-occlusive disease when slow flow is present. CTA results had a significant effect on patient clinical management.

http://www.ajnr.org/content/ajnr/26/5/1012.full.pdf

Intracranial vascular stenosis and occlusive disease: evaluation with CT angiography, MR angiography, and digital subtraction angiography.

Aneurysmal subarachnoid hemorrhage (SAH) has a 30-day mortality rate of 45%, with approximately half the survivors sustaining irreversible brain damage.1 On the basis of an annual incidence of 6 per 100 000, ≈15 000 Americans will have an aneurysmal SAH each year. Population-based incidence rates vary considerably from 6 to 16 per 100 000, with the highest rates reported from Japan and Finland.2 3 4 5 Approximately 5% to 15% of stroke cases are secondary to ruptured saccular aneurysms. Although the prevention of hemorrhage has been advocated as the most effective strategy aimed at lowering mortality rates,6 the optimal management of patients with unruptured intracranial aneurysms (UIAs) remains controversial. Management decisions require an accurate assessment of the risks of various treatment options compared with the natural history of the condition.

The natural history of UIAs and treatment outcomes are influenced by (1) patient factors, such as previous aneurysmal SAH, age, and coexisting medical conditions; (2) aneurysm characteristics, such as size, location, and morphology; and (3) factors in management, such as the experience of the surgical team and the treating hospital. These many influences have contributed to considerable variability in the reported risks for aneurysmal SAH and the treatment of UIAs. There are no prospective randomized trials of treatment interventions versus conservative management to date, and it is possible that no such studies will be carried out in the future.

According to a classification system suggested by Cook et al,7 randomized clinical trials with low likelihoods of false-positive and false-negative errors provide the highest level of evidence (level I) that can be applied to a clinical recommendation. Randomized trials with high likelihoods of false-negative and positive errors provide level II evidence. Level III evidence is generated with nonrandomized concurrent cohort comparisons between contemporaneous patients who did and …

/pdf/recommendations-for-the-management-of-patients-with-5f1pp2zzj3.pdf

Recommendations for the Management of Patients With Unruptured Intracranial Aneurysms A Statement for Healthcare Professionals From the Stroke Council of the American Heart Association

PURPOSE: To assess the accuracy of three-dimensional, Fourier transform, time-of-flight magnetic resonance (MR) angiography in the detection of intracranial steno-occlusive diseases. MATERIALS AND METHODS: One hundred thirty-one patients (62 male and 69 female patients, aged 6-77 years [mean, 53 years 8 months]) underwent MR and conventional angiography for evaluation of possible intracranial vascular disease. A total of 502 arteries were assessed. Eight projections and a collapsed image postprocessed by means of a maximum-intensity projection algorithm were reviewed by five observers in a blinded manner, with conventional angiography as the standard. RESULTS: A total of 32 steno-occlusive lesions were available for review. Receiver operating characteristic analysis from the pooled data revealed overall sensitivities of 85% and 88% and specificities of 96% and 97% for the internal carotid artery and the middle cerebral artery, respectively. CONCLUSION: MR angiography is useful as the primary diagnostic to...

Intracranial vascular stenosis and occlusion: diagnostic accuracy of three-dimensional, Fourier transform, time-of-flight MR angiography.

PURPOSE: To assess the accuracy of three-dimensional, Fourier transform, time-of-flight magnetic resonance (MR) angiography in the identification of intracranial aneurysms MATERIALS AND METHODS: MR angiograms of 126 patients (59 male and 67 female patients, aged 12-77 years) with various intracranial vascular lesions were evaluated Seventy-eight aneurysms, including 60 less than 5 mm in diameter, in 61 patients were depicted at conventional angiography Eight projection images, as well as one axial collapsed MR angiogram obtained with a maximum-intensity projection algorithm, were used for evaluation RESULTS: Sensitivity for the five observers ranged from 58% to 68% (mean, 63%) Higher sensitivity was achieved for anterior communicating and middle cerebral artery aneurysms, while that for internal carotid artery aneurysms was poor Sensitivities for small and medium aneurysms ranged from 50% to 60% (mean, 56%) and from 77% to 94% (mean, 86%), respectively CONCLUSION: MR angiography can depict intracra

Intracranial aneurysms: diagnostic accuracy of three-dimensional, Fourier transform, time-of-flight MR angiography.

PURPOSE To investigate whether obtaining axial source images from three-dimensional Fourier transform (3DFT) time-of-flight MR angiography improves the detection of intracranial vascular stenosis and occlusion if added to maximum-intensity projection (MIP) images. METHODS The angiograms of 103 patients who had MR angiography for evaluation of possible intracranial vascular disease were reviewed retrospectively in a quantitative and nonquantitative fashion. Diameters of vessels on MR angiograms were measured quantitatively by two reviewers using a magnifying loupe and compared with the results from conventional angiograms. Degrees of stenoocclusive disease were categorized into five classes; an artery with stenosis of 50% or greater was considered to be diseased. Another five observers also reviewed the MIP images with and without source images in a blinded fashion by means of nonquantitative visual inspection. RESULTS In all, 23 stenoocclusive lesions of 50% or greater were available for review. In the quantitative analysis, with MIP images alone, 14 (78%) of 18 moderate and severe stenoses and four (80%) of five occlusions were identified correctly. The addition of the source images increased the sensitivity to 100% for moderate and severe stenoses and to 100% for occluded vessels. In the visual inspection study, however, no statistically significant differences were found between interpretations of MIP images alone and those of MIP images in combination with source images. CONCLUSION In the quantitative study, interpretation of source images rather than MIP images reduced the tendency to overestimate stenosis seen with MR angiography and improved the sensitivity for detecting stenosis of 50% or greater. There was a discrepancy between the quantitative study and visual inspection. Experienced observers had a tendency to underestimate the degree of stenosis.

http://www.ajnr.org/content/ajnr/18/1/135.full.pdf

Intracranial vascular stenosis and occlusion: MR angiographic findings.

PURPOSE: To determine whether evaluation of source images from magnetic resonance (MR) angiography in addition to maximum-intensity projection (MIP) images improves the detection of aneurysms. MATERIALS AND METHODS: Conventional and MR angiography were performed in 193 patients with various intracranial vascular lesions or normal findings. Images were evaluated in a blinded manner. Two readings were performed 6 weeks apart by evaluating MIP images with and without source images. Results were evaluated with receiver operating characteristic analysis. RESULTS: Sensitivity for the detection of aneurysms increased slightly when source images were included. The detection rate of internal carotid artery aneurysms was most improved with the addition of source images. No statistically significant differences in performance were found between the readings with MIP images alone and with source images. CONCLUSION: Sensitivity may improve with combined reading of nonselective MIP and source images.

T O'Uchi

Papers

Intracranial vascular stenosis and occlusion: diagnostic accuracy of three-dimensional, Fourier transform, time-of-flight MR angiography.

Intracranial aneurysms: diagnostic accuracy of three-dimensional, Fourier transform, time-of-flight MR angiography.

Intracranial vascular stenosis and occlusion: MR angiographic findings.

Intracranial aneurysms: diagnostic accuracy of MR angiography with evaluation of maximum intensity projection and source images.