Incorporating algorithmic uncertainty into a clinical machine deep learning algorithm for urgent head CTs.

Machine learning (ML) algorithms to detect critical findings on head CTs may expedite patient management. Most ML algorithms for diagnostic imaging analysis utilize dichotomous classifications to determine whether a specific abnormality is present. However, imaging findings may be indeterminate, and algorithmic inferences may have substantial uncertainty. We incorporated awareness of uncertainty into an ML algorithm that detects intracranial hemorrhage or other urgent intracranial abnormalities and evaluated prospectively identified, 1000 consecutive noncontrast head CTs assigned to Emergency Department Neuroradiology for interpretation. The algorithm classified the scans into high (IC+) and low (IC-) probabilities for intracranial hemorrhage or other urgent abnormalities. All other cases were designated as No Prediction (NP) by the algorithm. The positive predictive value for IC+ cases (N = 103) was 0.91 (CI: 0.84-0.96), and the negative predictive value for IC- cases (N = 729) was 0.94 (0.91-0.96). Admission, neurosurgical intervention, and 30-day mortality rates for IC+ was 75% (63-84), 35% (24-47), and 10% (4-20), compared to 43% (40-47), 4% (3-6), and 3% (2-5) for IC-. There were 168 NP cases, of which 32% had intracranial hemorrhage or other urgent abnormalities, 31% had artifacts and postoperative changes, and 29% had no abnormalities. An ML algorithm incorporating uncertainty classified most head CTs into clinically relevant groups with high predictive values and may help accelerate the management of patients with intracranial hemorrhage or other urgent intracranial abnormalities.

Optimal brain MRI protocol for new neurological complaint.

BACKGROUND/PURPOSE: Patients with neurologic complaints are imaged with MRI protocols that may include many pulse sequences. It has not been documented which sequences are essential. We assessed the diagnostic accuracy of a limited number of sequences in patients with new neurologic complaints. METHODS: 996 consecutive brain MRI studies from patients with new neurological complaints were divided into 2 groups. In group 1, reviewers used a 3-sequence set that included sagittal T1-weighted, axial T2-weighted fluid-attenuated inversion recovery, and axial diffusion-weighted images. Subsequently, another group of studies were reviewed using axial susceptibility-weighted images in addition to the 3 sequences. The reference standard was the study's official report. Discrepancies between the limited sequence review and the reference standard including Level I findings (that may require immediate change in patient management) were identified. RESULTS: There were 84 major findings in 497 studies in group 1 with 21 not identified in the limited sequence evaluations: 12 enhancing lesions and 3 vascular abnormalities identified on MR angiography. The 3-sequence set did not reveal microhemorrhagic foci in 15 of 19 studies. There were 117 major findings in 499 studies in group 2 with 19 not identified on the 4-sequence set: 17 enhancing lesions and 2 vascular lesions identified on angiography. All 87 Level I findings were identified using limited sequence (56 acute infarcts, 16 hemorrhages, and 15 mass lesions). CONCLUSION: A 4-pulse sequence brain MRI study is sufficient to evaluate patients with a new neurological complaint except when contrast or angiography is indicated.

Diagnostic yield of delayed phase imaging in CT angiography of the head and neck: a retrospective study.

PURPOSE: To evaluate how often delayed images, obtained during neurovascular CTA, provide unique information relative to early phase imaging alone. MATERIALS AND METHODS: Informed consent was waived by the institutional review body for this study. Neurovascular CTAs from January through June 2009 were searched to identify those with delayed phase imaging. Reports were reviewed to identify cases where delayed images provided potentially unique information. The studies with potentially unique information were re-interpreted to determine if the information was indeed unique. RESULTS: 645 CTAs with delayed phase imaging were identified. There were 324 men and 310 women (median age 67 years; range 20-96 years). 59 studies (59/645: 9.1%) had findings on the delayed images. There were 13 cases with hemorrhage, with 4 showing progression on delayed views. Of the remaining 46 cases, 28 had occlusion of a vessel that did not reconstitute on the delayed images, 6 had occlusion of a vessel that did reconstitute on the delayed images, 7 had a string sign which was unchanged on the delayed views and 5 had no abnormal findings. Thus in 10 cases the findings were unique to the delayed images (10/645: 1.55%). Four showed active bleeding, three showed proximal occlusion with distal internal carotid filling from ophthalmic collaterals, two showed pial vessels filling distal to proximal MCA occlusion, and one showed retrograde vertebral artery filling due to subclavian steal. 95% confidence limits of the expected incidence of unique information from the delayed phase images are 0.6%-2.5%. CONCLUSION: Obtaining delayed phase imaging for neurovascular CTA should be an active decision and not the default protocol. This avoids imaging with little, if any value. If delayed images had not been obtained in our cohort, no detriment in patient management would have occurred.

Discriminating parathyroid adenoma from local mimics by using inherent tissue attenuation and vascular information obtained with four-dimensional CT: formulation of a multinomial logistic regression model.

PURPOSE: To identify a set of parameters, which are based on tissue enhancement and native iodine content obtained from a standardized triple-phase four-dimensional (4D) computed tomographic (CT) scan, that define a multinomial logistic regression model that discriminates between parathyroid adenoma (PTA) and thyroid nodules or lymph nodes. MATERIALS AND METHODS: Informed consent was waived by the institutional review board for this retrospective HIPAA-compliant study. Electronic medical records were reviewed for 102 patients with hyperparathyroidism who underwent triple-phase 4D CT and parathyroid surgery resulting in pathologically proved removal of adenoma from July 2010 through December 2011. Hounsfield units were measured in PTA, thyroid, lymph nodes, and aorta and were used to determine seven parameters characterizing tissue contrast enhancement. These were used as covariates in 10 multinomial logistic regression models. Three models with one covariate, four models with two covariates, and three models with three covariates were investigated. Receiver operating characteristic (ROC) analysis was performed to determine how well each model discriminated between adenoma and nonadenomatous tissues. Statistical differences between the areas under the ROC curves (AUCs) for each model pair were calculated, as well as sensitivity, specificity, accuracy, negative predictive value, and positive predictive value. RESULTS: A total of 120 lesions were found; 112 (93.3%) lesions were weighed, and mean and median weights were 589 and 335 mg, respectively. The three-covariate models were significantly identical (P > .65), with largest AUC of 0.9913 ± 0.0037 (standard error), accuracy of 96.9%, and sensitivity, specificity, negative predictive value, and positive predictive value of 94.3%, 98.3%, 97.1%, and 96.7%, respectively. The one- and two-covariate models were significantly less accurate (P < .043). CONCLUSION: A three-covariate multinomial logistic model derived from a triple-phase 4D CT scan can accurately provide the probability that tissue is PTA and performs significantly better than models using one or two covariates.

The Massachusetts General Hospital acute stroke imaging algorithm: an experience and evidence based approach.

The Massachusetts General Hospital Neuroradiology Division employed an experience and evidence based approach to develop a neuroimaging algorithm to best select patients with severe ischemic strokes caused by anterior circulation occlusions (ACOs) for intravenous tissue plasminogen activator and endovascular treatment. Methods found to be of value included the National Institutes of Health Stroke Scale (NIHSS), non-contrast CT, CT angiography (CTA) and diffusion MRI. Perfusion imaging by CT and MRI were found to be unnecessary for safe and effective triage of patients with severe ACOs. An algorithm was adopted that includes: non-contrast CT to identify hemorrhage and large hypodensity followed by CTA to identify the ACO; diffusion MRI to estimate the core infarct; and NIHSS in conjunction with diffusion data to estimate the clinical penumbra.

Accuracy of four-dimensional CT for the localization of abnormal parathyroid glands in patients with primary hyperparathyroidism.

PURPOSE: To investigate multiphase multidetector four-dimensional computed tomography (CT) as a technique to correctly localize abnormal parathyroid glands in patients with primary hyperparathyroidism. MATERIALS AND METHODS: Informed consent was waived by the institutional review body for this retrospective, chart review study. Radiology reports from four-dimensional CT and surgical notes were reviewed in 143 patients with primary hyperparathyroidism (35 men, 108 women; median ages, 58 and 60 years, respectively) who underwent parathyroid surgery between August 2004 and January 2007 and in whom four-dimensional CT predicted a single lesion. Accuracy of four-dimensional CT was stratified by patient and was determined separately for localization to the correct side and quadrant (upper and lower for each side), with surgical findings serving as standard of reference. RESULTS: In 143 patients, 148 abnormal parathyroid glands were found at surgery; 137 (93%) of these were weighed, with mean and median weights of 757 and 417 mg, respectively. Four-dimensional CT lateralized the abnormal glands with 93.7% accuracy (134 of 143). For localization according to quadrant, the accuracy was 86.6% (116 of 134). CONCLUSION: Four-dimensional CT has sufficiently high accuracy in presurgical localization to allow confident performance of unilateral parathyroidectomy in patients with sporadic primary hyperparathyroidism. The superior accuracy compared with that of ultrasonography and technetium 99m sestamibi scanning may be sufficient to allow four-dimensional CT to be used as the sole presurgical localization method.

Imaging characteristics of hyperfunctioning parathyroid adenomas using multiphase multidectector computed tomography: a quantitative and qualitative approach.

OBJECTIVE: The objective of the study was to characterize the enhancement pattern of hyperfunctioning parathyroid adenomas on multiphase multidetector computed tomography (CT) or 4-dimensional CT. METHODS: We retrospectively studied the enhancement patterns of 48 pathologically confirmed parathyroid adenomas with 4-dimensional CT, compliant with institutional review and the Health Insurance Portability and Accountability Act. Region-of-interest analysis was done at baseline and at arterial (25 seconds), venous (55 seconds), and delayed (85 seconds) enhancement phases over the adenoma and adjacent normal thyroid tissue. Qualitative and quantitative analysis was done. Discriminant functions were calculated using a multivariate logistic regression model, and receiver operating characteristic curves were measured. RESULTS: Adenomas are lower than thyroid in density, demonstrate avid early contrast enhancement, and show rapid wash-out of contrast. Adenomas and thyroid had baseline Hounsfield unit attenuations of 35 ± 11 and 94 ± 21 and enhancement percentage change from baseline to arterial of 493% ± 328% and 132% ± 148%, respectively (P < 0.0001 both). Quantitative analysis showed that these 2 measures of baseline density and the percentage change from baseline to arterial were the most powerful discriminatory features, with contrast wash-out from arterial peak to venous phase being a less powerful discriminator. Several discriminant functions were derived, the best of which was: X = 13.74 - (0.207 × baseline Hounsfield unit) - (0.006 × percent density change from baseline to arterial). X > 0.2 classifies tissue as parathyroid with high certainty (area under the receiver operating characteristic curve = 0.98; specificity, 0.938; sensitivity, 0.999). CONCLUSIONS: Parathyroid adenomas have a characteristic enhancement pattern that can be distinguished from thyroid tissue: the key diagnostic discriminators are baseline density, percentage change in density from baseline to arterial enhancement, and percentage decrease in density from arterial to venous phases.

Accuracy of 4-dimensional computed tomography in poorly localized patients with primary hyperparathyroidism.

BACKGROUND: Four-dimensional computed tomography (4D-CT) utilizes multiplanar images and perfusion characteristics to identify abnormal parathyroid glands. We assessed the role of 4D-CT in patients with inconclusive preoperative ultrasound and sestamibi localization studies. METHODS: Adult patients with primary hyperparathyroidism with negative or discordant standard imaging who underwent both localization with 4D-CT and operative intervention for curative intent were included. Patient characteristics, 4D-CT scan results compared with operative findings, and curative proportion were assessed. RESULTS: Of the 60 patients, 4D-CT accurately lateralized 73% and localized 60% of abnormal glands found at operation. Single candidate lesions (46/60) were confirmed at operation in 70%. When multiple lesions were identified on 4D-CT (14/60), accuracy dropped to 29% (P = .03). The accuracy of 4D-CT was not different between primary and reoperative cases (P = .79). Of the 8 patients with multigland disease diagnosed perioperatively, 5 had multiple candidate lesions noted on 4D-CT. In 94% (48/51) of patients, a >50% drop in intraoperative parathormone (IOPTH) level was achieved after resection and 87% (48/55) had long-term cure with a median follow-up of 221 days. CONCLUSION: 4D-CT identifies the more than half of abnormal parathyroids missed by traditional imaging and should be considered in cases with negative or discordant sestamibi and ultrasound. Bilateral exploration is warranted when multiple candidate lesions are reported on 4D-CT. Multigland disease remains a challenging entity.

Efficacy of 4D-CT preoperative localization in 2 patients with MEN 2A.

Multiple endocrine neoplasia type 2A (MEN2A) is an autosomal dominant syndrome that is associated with hyperparathyroidism in 20% to 30% of adult gene carriers. The appropriate surgical management of these patients remains in question. Approaches to this disease range from selective gland resection to a subtotal parathyroidectomy with or without autotransplantation. Despite surgical intervention, disease recurrence is problematic. Surgical management of patients found to have recurrence relies on localizing the anatomic location of the hyperfunctional gland(s). The primary imaging modality for localization of hyperfunctioning parathyroid glands is technetium 99m sestamibi single photon emission computed tomography (SPECT). Although sestamibi imaging has a sensitivity of 60% to 90%, specific anatomic detail is not always present by this imaging modality. Four-dimensional computed tomography (4D-CT) scans allow localization of ectopic parathyroid glands and autotransplanted parathyroid tissue, and they provide the anatomic detail necessary for decisions about appropriate surgical management. Another benefit of the 4D-CT scan is that enhancement characteristics, which are determined by contrast opacification of the hyperfunctional parathyroid tissue over 4 phases of the scan, correlate with metabolic activity. We recommend the use of 4D-CT scanning because of its capacity to identify hyperfunctional parathyroid glands and to provide anatomic information important in preoperative planning.

Parathyroid exploration in the reoperative neck: improved preoperative localization with 4D-computed tomography.

BACKGROUND: Reoperation for hyperparathyroidism (HPT) carries an increased risk for morbidity and failure to cure. Accurate preoperative localization minimizes operative risk but is often difficult to achieve in the reoperative setting. Four-dimensional computed tomography (4D-CT) is an emerging technique that uses functional parathyroid anatomy for precise preoperative localization. We evaluated 4D-CT as a tool for localization of hyperfunctioning parathyroid tissue in the reoperative setting. STUDY DESIGN: A prospective endocrine database was queried to identify 45 patients who underwent reoperative parathyroidectomy after preoperative localization using 4D-CT. The patients were categorized into 1 of 3 groups: group 1 included those who had previous neck surgery for non-HPT conditions; group 2 included those who had undergone a previously unsuccessful neck exploration for HPT; and group 3 included patients with HPT who had a previous neck exploration with resection of at least 1 hypercellular parathyroid. RESULTS: The sensitivity of 4D-CT for localization was 88% compared with 54% for sestamibi imaging. Four-dimensional CT more often correctly localized (p=0.0003) and lateralized (p=0.005) hyperfunctional parathyroid tissue than sestamibi did. Four-dimensional CT successfully localized hyperfunctional parathyroid tissue in 18 (82%) of 22 group 1 patients, 10 (91%) of 11 group 2 patients, and 8 (67%) of 12 group 3 patients. Three patients were lost to followup. At a mean followup of 9.8 months, 39 (93%) of 42 patients were surgically cured and 3 patients (7%; 2 in group 3) had persistent HPT. CONCLUSIONS: Four-dimensional-CT is an ideal tool for preoperative localization of hyperfunctioning parathyroid tissue in the reoperative setting. Localization and successful reoperation are most difficult in patients who have undergone an earlier operation that included resection of at least one hypercellular parathyroid suggesting multigland disease.

Current techniques used for the radiologic assessment of intracranial neoplasms.

CONTEXT: Radiologic studies are obtained for diagnostic and treatment planning purposes in the evaluation of patients with intracranial neoplasms. These studies are discussed at radiology/pathology working conferences and are often beneficial in the analysis of pathologic specimens for tissue diagnosis. Therefore, clinical pathologists should be aware of the current and evolving imaging techniques that are used in the radiologic assessment of intracranial neoplasms. OBJECTIVE: To describe the imaging techniques used in the assessment of intracranial neoplasms and provide current references. DATA SOURCES: We searched PubMed for articles published between 1995 and 2006 and also reviewed several textbooks on intracranial neoplasms, to prepare a discussion of basic modalities such as computed tomography (CT) and magnetic resonance (MR) imaging as well as advanced imaging techniques such as CT and MR angiography and venography, CT and MR perfusion, MR spectroscopy, functional MR imaging, and positron emission tomography. CONCLUSIONS: Knowledge of currently used imaging techniques for the assessment of intracranial neoplasms will assist the clinical pathologist in communications with neuroradiologists, surgeons, and referring clinicians. This review will also aid the pathologist in understanding the new and rapidly evolving imaging techniques that will likely become the standard of care in the future.

Improved preoperative planning for directed parathyroidectomy with 4-dimensional computed tomography.

BACKGROUND: Four-dimensional computed tomography (4D-CT) provides both functional and highly detailed anatomic information about parathyroid tumors. The purpose of this study was to compare 4D-CT with sestamibi imaging and ultrasonography as methods for the accurate preoperative localization of hyperfunctioning parathyroid glands before parathyroidectomy. METHODS: A study of 75 patients with primary hyperparathyroidism was performed at a tertiary-care institution. Sestamibi imaging, ultrasonography, and 4D-CT were performed on each patient preoperatively. Results of the imaging studies were compared with operative findings, pathologic data, and biochemical measurements to assess the sensitivity and specificity of each of the imaging modalities. RESULTS: 4D-CT demonstrated improved sensitivity (88%) over sestamibi imaging (65%) and ultrasonography (57%), when the imaging studies were used to localize (lateralize) hyperfunctioning parathyroid glands to 1 side of the neck. Moreover, when used to localize parathyroid tumors to the correct quadrant of the neck (ie, right inferior, right superior, left inferior, or left superior), the sensitivity of 4D-CT (70%) was significantly higher than sestamibi imaging (33%) and ultrasonography (29%). CONCLUSION: 4D-CT provides significantly greater sensitivity than sestamibi imaging and ultrasonography for precise (quadrant) localization of hyperfunctioning parathyroid glands. This allows improved preoperative planning, particularly for the case of reoperation. In addition to the data that are provided, we present a novel classification scheme for use in parathyroid localization.

Comparison of image quality between conventional and low-dose nonenhanced head CT.

BACKGROUND AND PURPOSE: Increasing use of CT for evaluating neurologic disease may expose patients to considerable levels of ionizing radiation. We compared the image quality of low-mAs head CT scans with that of conventional nonenhanced scans. METHODS: Conventional head CT scans were obtained in 20 patients (all >65 years with history of non-CNS malignancy) by using a multidetector technique: 170 mA and 1-second scanning time (ie, 170 mAs), 140 kVp, table speed of 7.5 mm per rotation, pitch of 0.75, section thickness of 5 mm, and field of view of 25 mm. A limited volume helical data acquisition covering four 5-mm-thick images was obtained by using 90 mAs but otherwise the same parameters. Three neuroradiologists visually rated the resulting images for quality in a blinded comparison. Representative 1- to 4-mm(2) regions of interest were chosen in gray matter and white matter locations. Conspicuity and the contrast-to-noise ratio were analyzed. Statistical comparisons were done by using the Student t test. RESULTS: Mean gray matter conspicuity was not significantly different between the 170- and 90-mAs groups (0.39 +/- 0.19 vs 0.41 +/- 0.03, P =.32). Mean gray matter contrast-to-noise ratio was approximately 22% higher with 170 mAs than with 90 mAs (1.77 +/- 0.52 vs 1.39 +/- 0.38, P =.005). All 90-mAs images were rated as having slightly greater image noise than the 170-mAs scans but with sufficient perceived resolution. CONCLUSION: Although 90-mAs head CT images were moderately noisier than 170-mAs images, they were rated as having acceptable diagnostic quality.

Whole-brain CT perfusion measurement of perfused cerebral blood volume in acute ischemic stroke: probability curve for regional infarction.

PURPOSE: To determine the probability curve for regional cerebral infarction as a function of percentage normalized perfused cerebral blood volume (pCBV) in patients with acute ischemic stroke. MATERIALS AND METHODS: The authors retrospectively analyzed whole-brain computed tomographic (CT) perfusion scans from 28 patients with acute stroke (<6 hours) due to major arterial occlusion, without intracranial hemorrhage. Each patient had a positive follow-up CT scan 1-4 days later, without interval thrombolysis. Normalized pCBV, expressed as a percentage of contralateral normal brain pCBV, was determined in the core infarction and in regions just inside and outside the boundary between infarcted and noninfarcted brain. These regions were dichotomized into infarcted (core and inner band) and noninfarcted (outer band) categories. Logistic regression analysis was then used to create a reference curve of probability of infarction as a function of percentage normalized pCBV. RESULTS: Normalized pCBV values in the core, inner band, and outer band were 24.5% +/- 2.3, 36.3% +/- 2.4, and 72.1% +/- 2.4, with corresponding probabilities of infarction of .99, .96, and .11. The normalized pCBV at which the probability of survival reached .5 was 58.0% +/- 0.5. Sensitivity, specificity, and accuracy of the reference probability curve were 90.5% (209 of 231), 89.5% (212 of 237), and 90.0% (421 of 468), respectively. Negative and positive predictive values were 90.6% (212 of 234) and 89.3% (209 of 234), respectively. R2 was 0.73, and differences in perfusion between core and inner and outer bands were highly significant (P <.0001). CONCLUSION: A probability of infarction curve can help predict the likelihood of infarction as a function of percentage normalized pCBV.

Multi-detector row CT: radiation dose characteristics.

PURPOSE: To determine the dose characteristics of multi-detector row computed tomography (CT) and to provide tabulated dose values and rules of thumb that assist in minimizing the radiation dose at multi-detector row CT. MATERIALS AND METHODS: Weighted CT dose index (CTDI100w) values were obtained from three multi-detector row CT scanners (LightSpeed; GE Medical Systems, Milwaukee, Wis) for both head and body CT modes by using standard CT-dose phantoms. The CTDI100w was determined as a function of x-ray tube voltage (80, 100, 120, 140 kVp), tube current (range, 50-380 mA), tube rotation time (0.5-4.0 seconds), radiation profile width (RPW) (5, 10, 15, 20 mm), and acquisition mode (helical high-quality and high-speed modes and axial one-, two-, and four-section modes). Statistical regression was performed to characterize the relationships between CTDI100w and various technique factors. RESULTS: The CTDI100w (milligray) increased linearly with tube current: in head mode, CTDI100w = (0.391 mGy/mA +/- 0.004) x tube current (milliampere) (r2 = 0.999); in body mode, CTDI100w = (0.162 mGy/mA +/- 0.002) x tube current (milliampere) (r2 = 0.999). The CTDI100w increased linearly with rotation time: in head mode, CTDI100w = (34.7 mGy/sec +/- 0.2) x rotation time (seconds) (r2 = 1.0); in body mode, CTDI100w = (13.957 mGy/sec +/- 0.005) x rotation time (seconds) (r2 = 1.0). The relationship of normalized CTDI100w (milligrays per 100 mAs) with tube voltage followed a power law: in head mode, CTDI100w = (0.00016 mGy/100 mAs. kVp +/- 0.00007) x (tube voltage)(2.5+/-0.1) (r2 = 0.997); in body mode, CTDI100w = (0.000012 mGy/100 mAs. kVp +/- 0.000007) x (tube voltage)(2.8+/-0.1) (r2 = 0.996). In all scanning modes, CTDI100w decreased when RPW increased. CTDI100w was 10% higher in head mode and 13% lower in body mode compared with the value suggested by the manufacturer, which is displayed at the scanner console. When deposited power exceeded 24 kW, CTDI100w increased by 10% as a result of use of the large focal spot. CONCLUSION: The authors provide a set of tables of radiation dose as a function of imaging protocol to facilitate implementation of radiation dose-efficient studies.

Assessing tissue viability with MR diffusion and perfusion imaging.

BACKGROUND AND PURPOSE: Diffusion- (DW) and perfusion-weighted (PW) MR imaging reflect neurophysiologic changes during stroke evolution. We sought to determine parameters that distinguish regions of brain destined for infarction from those that will survive despite hypoperfusion. METHODS: DW and PW images were obtained in 30 patients at 1-12 hours after symptom onset. Relative cerebral blood volume (rCBV), flow (rCBF), mean transit time (MTT), apparent diffusion coefficient (ADC), DW image signal intensity, and fractional anisotropy (FA) lesion-contralateral normal region ratios were obtained in the following regions: 1) infarct core with hyperintensity on DW image, abnormality on rCBF and MTT images, and follow-up abnormality; 2) infarcted penumbra with normal DW image, abnormal rCBF and MTT images, and follow-up abnormality; and 3) hypoperfused tissue that remained viable, with normal DW image, abnormal rCBF and MTT images, and normal follow-up. RESULTS: rCBF ratios for regions 1, 2, and 3 were 0.32 +/- 0.11, 0.46 +/- 0.13, and 0.58 +/- 0.12, respectively, and were significantly different. DW image intensity and ADC ratios were significantly different among all regions, but were more similar than rCBF ratios. rCBV and FA ratios were not significantly different between regions 2 and 3. No MTT ratios were significantly different. No region of interest with an rCBF ratio less than 0.36, an rCBV ratio less than 0.53, an ADC ratio less than 0.85, a DW image intensity ratio greater than 1.23, or an FA ratio greater than 1.10 remained viable. No region of interest with an rCBF ratio greater than 0.79 infarcted. CONCLUSIONS: Differences among mean ratios of three regions investigated were greatest for the rCBF ratio. The rCBF ratio may be the most useful parameter in differentiating viable tissue that is likely to infarct without intervention, from tissue that will survive despite hypoperfusion. ADC, DW intensity, FA, and rCBV ratios may provide adjunctive information.