Eye Tracking for Deep Learning Segmentation Using Convolutional Neural Networks.

Deep learning with convolutional neural networks (CNNs) has experienced tremendous growth in multiple healthcare applications and has been shown to have high accuracy in semantic segmentation of medical (e.g., radiology and pathology) images. However, a key barrier in the required training of CNNs is obtaining large-scale and precisely annotated imaging data. We sought to address the lack of annotated data with eye tracking technology. As a proof of principle, our hypothesis was that segmentation masks generated with the help of eye tracking (ET) would be very similar to those rendered by hand annotation (HA). Additionally, our goal was to show that a CNN trained on ET masks would be equivalent to one trained on HA masks, the latter being the current standard approach. Step 1: Screen captures of 19 publicly available radiologic images of assorted structures within various modalities were analyzed. ET and HA masks for all regions of interest (ROIs) were generated from these image datasets. Step 2: Utilizing a similar approach, ET and HA masks for 356 publicly available T1-weighted postcontrast meningioma images were generated. Three hundred six of these image + mask pairs were used to train a CNN with U-net-based architecture. The remaining 50 images were used as the independent test set. Step 1: ET and HA masks for the nonneurological images had an average Dice similarity coefficient (DSC) of 0.86 between each other. Step 2: Meningioma ET and HA masks had an average DSC of 0.85 between each other. After separate training using both approaches, the ET approach performed virtually identically to HA on the test set of 50 images. The former had an area under the curve (AUC) of 0.88, while the latter had AUC of 0.87. ET and HA predictions had trimmed mean DSCs compared to the original HA maps of 0.73 and 0.74, respectively. These trimmed DSCs between ET and HA were found to be statistically equivalent with a p value of 0.015. We have demonstrated that ET can create segmentation masks suitable for deep learning semantic segmentation. Future work will integrate ET to produce masks in a faster, more natural manner that distracts less from typical radiology clinical workflow.

Pathophysiology and Imaging Findings of COVID-19 Infection: An Organ-system Based Review.

BACKGROUND: COVID-19 commonly presents with upper respiratory symptoms; however, studies have shown that SARS-CoV-2 infection affects multiple organ systems. Here, we review the pathophysiology and imaging characteristics of SARS-CoV-2 infection in organ systems throughout the body and explore commonalities. OBJECTIVE: Familiarity with the underlying pathophysiology and imaging characteristics is essential for the radiologist to recognize these findings in patients with COVID-19 infection. Though pulmonary findings are the most prevalent presentation, COVID-19 may have multiple manifestations and recognition of the extrapulmonary manifestations is especially important because of the potential serious and long-term effects of COVID-19 on multiple organ systems.

Local Glioma Cells Are Associated with Vascular Dysregulation.

BACKGROUND AND PURPOSE: Malignant glioma is a highly infiltrative malignancy that causes variable disruptions to the structure and function of the cerebrovasculature. While many of these structural disruptions have known correlative histopathologic alterations, the mechanisms underlying vascular dysfunction identified by resting-state blood oxygen level-dependent imaging are not yet known. The purpose of this study was to characterize the alterations that correlate with a blood oxygen level-dependent biomarker of vascular dysregulation. MATERIALS AND METHODS: Thirty-two stereotactically localized biopsies were obtained from contrast-enhancing (n = 16) and nonenhancing (n = 16) regions during open surgical resection of malignant glioma in 17 patients. Preoperative resting-state blood oxygen level-dependent fMRI was used to evaluate the relationships between radiographic and histopathologic characteristics. Signal intensity for a blood oxygen level-dependent biomarker was compared with scores of tumor infiltration and microvascular proliferation as well as total cell and neuronal density. RESULTS: Biopsies corresponded to a range of blood oxygen level-dependent signals, ranging from relatively normal (z = -4.79) to markedly abnormal (z = 8.84). Total cell density was directly related to blood oxygen level-dependent signal abnormality (P = .013, R2 = 0.19), while the neuronal labeling index was inversely related to blood oxygen level-dependent signal abnormality (P = .016, R2 = 0.21). The blood oxygen level-dependent signal abnormality was also related to tumor infiltration (P = .014) and microvascular proliferation (P = .045). CONCLUSIONS: The relationship between local, neoplastic characteristics and a blood oxygen level-dependent biomarker of vascular function suggests that local effects of glioma cell infiltration contribute to vascular dysregulation.

A Multiparametric Model for Mapping Cellularity in Glioblastoma Using Radiographically Localized Biopsies.

BACKGROUND AND PURPOSE: The complex MR imaging appearance of glioblastoma is a function of underlying histopathologic heterogeneity. A better understanding of these correlations, particularly the influence of infiltrating glioma cells and vasogenic edema on T2 and diffusivity signal in nonenhancing areas, has important implications in the management of these patients. With localized biopsies, the objective of this study was to generate a model capable of predicting cellularity at each voxel within an entire tumor volume as a function of signal intensity, thus providing a means of quantifying tumor infiltration into surrounding brain tissue. MATERIALS AND METHODS: Ninety-one localized biopsies were obtained from 36 patients with glioblastoma. Signal intensities corresponding to these samples were derived from T1-postcontrast subtraction, T2-FLAIR, and ADC sequences by using an automated coregistration algorithm. Cell density was calculated for each specimen by using an automated cell-counting algorithm. Signal intensity was plotted against cell density for each MR image. RESULTS: T2-FLAIR (r = -0.61) and ADC (r = -0.63) sequences were inversely correlated with cell density. T1-postcontrast (r = 0.69) subtraction was directly correlated with cell density. Combining these relationships yielded a multiparametric model with improved correlation (r = 0.74), suggesting that each sequence offers different and complementary information. CONCLUSIONS: Using localized biopsies, we have generated a model that illustrates a quantitative and significant relationship between MR signal and cell density. Projecting this relationship over the entire tumor volume allows mapping of the intratumoral heterogeneity in both the contrast-enhancing tumor core and nonenhancing margins of glioblastoma and may be used to guide extended surgical resection, localized biopsies, and radiation field mapping.

SPECT study of visual fixation in schizophrenia and comparison subjects.

BACKGROUND: The consistent association of impaired eye movements and schizophrenia suggests a relationship between the neurobiology of the illness and visual pursuit systems. Visual fixation (VF), an eye "movement" task at zero velocity, is the simplest such abnormality in schizophrenia patients and their relatives. METHODS: We used a VF task for a functional imaging study. Six neuroleptic-free schizophrenia patients and eight gender and mean age matched comparison subjects had SPECT scans with 20 mCi of Tc99-HMPAO, during VF on a simple blue line intersection. MEDX data saved in ANALYZE format for SPM 95 was used to generate paired t-test statistical data for display in Talairach space, with rCBF changes given as Z-scores. RESULTS: Patients, compared to controls, had increased rCBF in both the parahippocampal gyrus (bilaterally) and in the right fusiform gyrus. They had decreased rCBF in the left frontal cortex, including medial and superior frontal gyri and anterior cingulate. Overall, compared to controls, patients had medial temporal lobe hyperperfusion along with left prefrontal hypoperfusion. CONCLUSIONS: These findings are consistent with the hypothesized imbalance between the medial temporal and frontal lobes that is postulated for schizophrenia. It was of interest that the relative rCBF differences between schizophrenia patients and controls in this small sample were observable with this cognitively non-demanding visual fixation task.

SPECT imaging of odor identification in schizophrenia.

Deficits in olfactory identification, despite normal odor perception, are found in some neuropsychiatric disorders, including schizophrenia. We examined if regional cerebral blood flow (rCBF) differed between schizophrenia patients and controls during odor identification, hypothesizing that these brain regions could be relevant to odor identification impairments. Eight schizophrenia and eight comparison subjects provided a baseline (picture identity matching) and activation (odor identification) SPECT scan, obtained using 99mTc-HMPAO in a low dose/high dose design. Six patients and seven controls had analyzable data. MEDX data saved in ANALYZE format for SPM 95 generated paired t-test statistical data for display in Talairach space, with rCBF changes given as Z-scores. There was no schizophrenia vs. control group difference in rCBF for the baseline picture-matching test. For odor identification, schizophrenia patients had a hypometabolic right-sided cortical region that included the frontal lobe Broca's area, superior temporal lobe, and supramarginal and angular gyri. Post hoc within-group contrasts of picture-matching vs. odor identification showed that the controls significantly increased rCBF in the right-sided inferior temporal fusiform gyrus, and bilateral hippocampi and visual association areas for the odor test. The schizophrenia group showed no rCBF differences for picture-matching compared to odor identification. Patients showed significant hypometabolism in right-sided cortical areas for odor identification. They also failed to show increased rCBF in the hippocampus and visual association area, as seen in controls for odor identification compared to picture-matching. These regions may be unique to schizophrenia or have broader implications for olfactory memory retrieval.

Semiautomated volumetric measurement on postcontrast MR imaging for analysis of recurrent and residual disease in glioblastoma multiforme.

BACKGROUND AND PURPOSE: A limitation in postoperative monitoring of patients with glioblastoma is the lack of objective measures to quantify residual and recurrent disease. Automated computer-assisted volumetric analysis of contrast-enhancing tissue represents a potential tool to aid the radiologist in following these patients. In this study, we hypothesize that computer-assisted volumetry will show increased precision and speed over conventional 1D and 2D techniques in assessing residual and/or recurrent tumor. MATERIALS AND METHODS: This retrospective study included patients with native glioblastomas with MR imaging performed at 24-48 hours following resection and 2-4 months postoperatively. 1D and 2D measurements were performed by 2 neuroradiologists with Certificates of Added Qualification. Volumetry was performed by using manual segmentation and computer-assisted volumetry, which combines region-based active contours and a level set approach. Tumor response was assessed by using established 1D, 2D, and volumetric standards. Manual and computer-assisted volumetry segmentation times were compared. Interobserver correlation was determined among 1D, 2D, and volumetric techniques. RESULTS: Twenty-nine patients were analyzed. Discrepancy in disease status between 1D and 2D compared with computer-assisted volumetry was 10.3% (3/29) and 17.2% (5/29), respectively. The mean time for segmentation between manual and computer-assisted volumetry techniques was 9.7 minutes and <1 minute, respectively (P < .01). Interobserver correlation was highest for volumetric measurements (0.995; 95% CI, 0.990-0.997) compared with 1D (0.826; 95% CI, 0.695-0.904) and 2D (0.905; 95% CI, 0.828-0.948) measurements. CONCLUSIONS: Computer-assisted volumetry provides a reproducible and faster volumetric assessment of enhancing tumor burden, which has implications for monitoring disease progression and quantification of tumor burden in treatment trials.

Perilesional blood flow and edema formation in acute intracerebral hemorrhage: a SPECT study.

BACKGROUND AND PURPOSE: Secondary brain injury and edema formation contribute significantly to morbidity and mortality after intracerebral hemorrhage (ICH). The pathogenesis of this process is poorly understood. We sought to characterize alterations in perilesional blood flow that occur during the acute phase of ICH and to determine whether progressive enlargement of edema surrounding ICH is related to increased or decreased perfusion. METHODS: We performed paired consecutive CT and 99mTc-hexamethylpropylenamine oxime single-photon emission computed tomography (SPECT) scans during the acute (mean, 18 hours) and subacute (mean, 72 hours) phase of ICH in 23 patients. Hematoma and edema volumes were traced and calculated from CT images. SPECT-derived hypothetical flow deficit volumes (FDV) around each hematoma were calculated by measuring a "zero-flow" volume within a large perilesional region of interest (based on percent tracer count loss compared with the contralateral side) and subtracting the corresponding ICH volume. Patients with significant midline shift (>5 mm) or global blood flow reduction were excluded from the analysis. RESULTS: ICH volume (18 mL) did not change, mean edema volume increased by 36% (from 19 to 25 mL, P<0.0001), and mean FDV decreased by 55% (from 14 to 6 mL, P=0.0004) between the acute and subacute phases. Edema volume on the second CT scan correlated positively with FDV on the first SPECT scan (Spearman's p=0.48, P=0.02), and with the volume of reperfused perilesional tissue (FDVacute-FDVsubacute) (Spearman's p=0.41, P=0.05). Perilesional edema on CT always corresponded topographically with perfusion deficits on SPECT. In 4 patients, delayed focal hyperemia was identified in more peripheral cortical regions, but these areas appeared normal on CT. CONCLUSIONS: Perilesional blood flow normalizes from initially depressed levels as edema forms during the first 72 hours after ICH, and the eventual extent of edema correlates with the volume of reperfused tissue. These results suggest that the potential for perilesional ischemia is highest in the earliest hours after ICH onset and implicate reperfusion injury in the pathogenesis of perihematoma edema formation.