A novel stacked generalization of models for improved TB detection in chest radiographs.

Chest x-ray (CXR) analysis is a common part of the protocol for confirming active pulmonary Tuberculosis (TB). However, many TB endemic regions are severely resource constrained in radiological services impairing timely detection and treatment. Computer-aided diagnosis (CADx) tools can supplement decision-making while simultaneously addressing the gap in expert radiological interpretation during mobile field screening. These tools use hand-engineered and/or convolutional neural networks (CNN) computed image features. CNN, a class of deep learning (DL) models, has gained research prominence in visual recognition. It has been shown that Ensemble learning has an inherent advantage of constructing non-linear decision making functions and improve visual recognition. We create a stacking of classifiers with hand-engineered and CNN features toward improving TB detection in CXRs. The results obtained are highly promising and superior to the state-of-the-art.

A comparative study of thoracic radiation doses from 64-slice cardiac CT.

The goal of this study was to measure radiation doses for 64-slice cardiac CT angiography studies and to study the dose-savings features of these CT scanners. This was done using various phantoms. These radiation doses were compared with those from typical helical body CT scans, fluoroscopy cardiac catheterization studies and mammography examinations. Radiation measurements were made with a CT ionization detector and a solid state dosimeter. A GE 64-slice Lightspeed VCT and a Siemens Somatom Sensation 64 CT were used to scan a standard 32 cm acrylic phantom and an anthropomorphic phantom. Data were collected in axial and various gated cardiac helical modes. Organ doses and the effective doses were calculated from the measurements. In gated CT cardiac mode with the 32 cm acrylic phantom, the measured radiation doses per study were generally three to seven times greater than those from typical body helical CT examinations; the range depended upon selectable scan parameters. With the anatomical phantom, the surface doses in the anteroposterior (AP) plane were typically 20-60% higher than those measured using the 32 cm phantom. The lateral surface doses were -4% to +15%. These results can be attributed to the shorter AP dimension and the air in the lungs. The CT skin entrance radiation doses were 80-90% less than diagnostic cardiac catheterization studies, and organ doses were similar. Because 64-slice cardiac gated CT uses pitches equal to 0.20-0.27 and high mAs values, the patient radiation doses are appreciably higher than in routine body CT examinations. The female breast, which could receive a radiation dose 10-30 times that received from mammography screening, is an organ of particular concern.

2001 Joseph E. and Nancy O. Whitley Award. Teaching noninterpretive skills to radiology residents: a collaborative effort between the American College of Radiology and the Association of Program Directors in Radiology.

RATIONALE AND OBJECTIVES: The authors' purpose was to develop, implement, and evaluate a series of videotapes for teaching noninterpretive skills to radiology residents. MATERIALS AND METHODS: An Association of Program Directors in Radiology/American College of Radiology working group developed a series of seven videotapes to teach residents job search and contracting skills; business aspects of radiology; American College of Radiology standards, accreditation programs, and appropriateness criteria; critical thinking skills; ethics; service orientation and interpersonal skills; and medical organizational politics. Residents viewed the videotapes, evaluated the learning experience, and completed both pre- and posttests. RESULTS: From 85 to 264 residents completed each of six pre- and posttests. Mean total pretest scores ranged from 55.0% to 76.9% and improved significantly (P < .05) to 73.8% to 94.9% on posttests. Each videotape was evaluated by 90-336 residents using a Likert-type scale, with a rating of 5 representing the most positive response. The proportion of residents who gave each tape an overall rating of either 4 or 5 varied from 75.9% to 95.9%. CONCLUSION: Significant resident learning occurred as a result of viewing the videotapes. Program director support of the learning experience was strong, and use of the videotapes as a teaching modality received strong support by residents. Local moderators and question-and-answer sessions would optimize the learning experience.

Early retardation of 99mTc-DTPA radioaerosol transalveolar clearance in irradiated canine lung.

UNLABELLED: The alteration of 99mTc-labeled diethylenetriaminepentaacetic acid (DTPA) transalveolar clearance in an initial phase of radiation lung injury was experimentally investigated. METHODS: Fourteen dogs were irradiated to the hemithorax with a single dose of 20 Gy. A DTPA radioaerosol study was performed before irradiation and on day 12 after irradiation. On day 14, the DTPA study was repeated again, with seven animals undergoing the study after inhalation of an aerosolized synthetic surfactant. The penetration index (P.I.) and clearance half-time (T(1/2)) of DTPA were measured in each lung. To evaluate the changes in lung surfactant after irradiation, alveolar lipids were stained in the resected lungs (n = 14), and the amounts of alveolar surfactant phospholipid and protein were measured by a bronchoalveolar lavage study in another six irradiated dogs. RESULTS: In all of the 14 irradiated animals, DTPA radioaerosol distributed uniformly throughout the lungs without significant changes in P.I. The T(1/2) values in irradiated lungs were significantly prolonged compared with the matched baseline values and those in nonirradiated lungs (P < 0.05 and 0.001, respectively). The aerosolized synthetic surfactant retarded the DTPA clearance both in the irradiated and in the nonirradiated lungs (P < 0.001) without significant changes in P.I. The histologic and bronchoalveolar lavage studies revealed an increase of alveolar surfactant materials in the irradiated lungs without substantial histologic changes in the alveolar structures. CONCLUSION: DTPA transalveolar clearance was retarded soon after irradiation. Increased alveolar surfactant may be partly responsible for this retarded DTPA clearance because the aerosolized synthetic surfactant also prolonged the clearance in nonirradiated lungs. A DTPA clearance test is sensitive for the early detection of radiation lung injury and seems helpful for clarifying the association of epithelial integrity changes and lung surfactant in radiation lung injury.

Rapid in vivo monitoring of chemotherapeutic response using weighted sodium magnetic resonance imaging.

A novel pulse sequence strategy uses sodium magnetic resonance imaging to monitor the response to chemotherapy of mouse xenograft tumors propagated from human prostate cancer cell lines. An inversion pulse suppresses sodium with long longitudinal relaxation times, weighting the image toward intracellular sodium nuclei. Comparing these weighted sodium images before and 24 h after administration of antineoplastics, we measured a 36 +/- 4% (P < 0.001; n = 16) increase in signal intensity. Experiments with these same drugs and cells, treated in culture, detected a significant intracellular sodium elevation (10-20 mM) using a ratiometric fluorescent dye. Flow cytometry studies showed that this elevation preceded cell death by apoptosis, as determined by fluorescent end-labeling of apoptotic nuclei or Annexin V binding. Histopathology on formalin-fixed sections of explanted tumors confirmed that drug administration reduces proliferation (2.2 versus 8.6 mitotic figures per high power field; P < 0.0001), an effect that inversely correlates with the sodium magnetic resonance image response on a tumor-to-tumor basis (P < 0.02; n = 10). Morphological features, such as central zones of nonviable cells, rims of active apoptosis, and areas of viable tumor, could be distinguished by comparing weighted and unweighted images. Advantages of this sodium imaging technique include rapid determination of drug efficacy, improved diagnosis of lesions, ease of coregistration with high resolution proton magnetic resonance imaging, and absence of costly or toxic reagents.

Radioaerosol scintigraphy in infants and children born to mothers with HIV disease. Pediatric Pulmonary and Cardiovascular Complications (of Vertically Transmitted Human Immunodeficiency Virus) Study Group.

PURPOSE: To determine the usefulness of technetium 99m diethyltriaminepentacetic acid (DTPA) radioaerosol inhalation-clearance scintigraphy for early detection of pulmonary complications of human immunodeficiency virus (HIV) disease in children. MATERIALS AND METHODS: A total of 301 studies were performed in 132 HIV-positive children (group 1; mean age, 46.6 months). In children born to HIV-positive mothers (group 2), 273 studies were performed in 160 children who eventually were proved to be HIV negative (mean age, 10.3 months), and 80 studies were performed in 47 HIV-positive children (mean age, 15.6 months). Radioaerosol studies were performed by using commercially available radioaerosol nebulizers. Pulmonary clearance half-time was measured by using conventional gamma camera computer systems. Radioaerosol results were correlated with indexes of pulmonary health and function. RESULTS: The HIV-negative, group 2 children had a mean radioaerosol clearance half-time (58.1 minutes; 162 studies in 108 children) similar to that reported in healthy adults. Group 1 children with pulmonary involvement exhibited a faster mean clearance half-time (28.6 minutes) than did children without evidence of pulmonary involvement from either group 1 or group 2 (P < .05). A faster pulmonary clearance rate did not simply reflect the presence of chest disease that also was detectable on radiographs (P = .3). CONCLUSION: Quantitative DTPA radioaerosol clearance studies may provide useful information about pulmonary involvement in selected children with HIV disease.

Effect of inhaled surfactant on pulmonary deposition and clearance of technetium-99m-DTPA radioaerosol.

UNLABELLED: To establish the effect of an aerosolized synthetic surfactant (Exosurf) on pulmonary 99mTc-diethylenetriamine pentaacetic acid (DTPA) aerosol deposition and clearance, radioaerosol studies were performed at varying times and under varying conditions after surfactant inhalation in canine lungs. METHODS: Twenty-three dogs had a baseline 99mTc-DTPA study; 2 days later the study was repeated after inhalation of a 1.5-ml/kg dose of Exosurf aerosolized by an ultrasonic nebulizer. The clearance half-time (T1/2) of 99mTc-DTPA from each lung was measured at different times (from 10 min to 3 hr) after Exosurf inhalation. For comparison, five animals had a 99mTc-DTPA study 10 min after inhalation of the same dose of saline as Exosurf. An additional five animals inhaled a 99mTc-DTPA-Exosurf mixture to investigate the distribution of Exosurf. RESULTS: Technetium-99m-DTPA distributed uniformly without significant changes in penetration indexes before and after inhalation of Exosurf and the 99mTc-DTPA-Exosurf mixture. After Exosurf inhalation, 99mTc-DTPA clearance at 10 min (T1/2; 35.6 +/- 8.7 min; n = 6) and 40 min (29.4 +/- 6.3 min; n = 4) was significantly prolonged compared with the matched baseline values (24.7 +/- 6.4 min, p < 0.0001; and 21.7 +/- 8.9 min, p = 0.01, respectively). However, later clearance times were not prolonged. By contrast, after saline inhalation, 99mTc-DTPA distributed inhomogeneously, and clearance times T1/2) were not altered from the matched baseline values. CONCLUSION: Aerosolized Exosurf distributes homogeneously in the lungs. Exosurf initially retards 99mTc-DTPA aerosol clearance, but 99mTc-DTPA transalveolar clearance returns to baseline rates within 1-2 hr. Technetium-99m-DTPA aerosol clearance measurements can be used to monitor the effect of inhaled Exosurf on pulmonary epithelial integrity.

Association between right ventricular function and perfusion abnormalities in hemodynamically stable patients with acute pulmonary embolism.

BACKGROUND/OBJECTIVES: Patients presenting with acute pulmonary embolism associated with hemodynamic compromise exhibit right ventricular enlargement and dysfunction on transthoracic echocardiogram. However, the degree of echocardiographic abnormalities among hemodynamically stable patients without preexisting cardiopulmonary disease during the acute stage of pulmonary embolism, and following treatment, is unknown. Therefore, this study was designed to assess the extent of right ventricular abnormalities detected on transthoracic echocardiogram in patients following acute pulmonary embolism and during treatment with anticoagulation or vena caval interruption. The extent of pulmonary vascular obstruction and complication rate on follow-up were also assessed. DESIGN/INTERVENTIONS: Sixty-four consecutive hemodynamically stable patients without preexisting known cardiopulmonary disorder presenting with acute pulmonary embolism and undergoing treatment with anticoagulation or inferior vena caval interruption were studied. All subjects underwent a two-dimensional transthoracic echocardiogram within 24 h of diagnosis. The degree of perfusion abnormality on lung scan was quantified. Twenty-six patients underwent follow-up echocardiogram and lung scan at 6 weeks. The echocardiographic findings were compared with those obtained from a group of normal control subjects matched for gender and age. RESULTS: Although the mean right ventricular end-diastolic areas did not differ (21.9+/-5.2 cm2 vs 20.1+/-2.9 cm2 for control subjects; p=not significant), the right ventricular end-systolic area was larger in comparison to our series of control subjects (14.6+/-5.1 cm2 vs 11.7+/-2.0 cm2; p=0.025). Fractional right ventricular area change was reduced in the patient group compared with the control subjects (34.3+/-9.0% vs 41.3+/-7.0%; p=0.003). The extent of right ventricular end-systolic area enlargement and decrease in fractional area change did not correlate with the degree of pulmonary vascular obstruction. Patients who were restudied at 6 weeks showed minimal improvement in echocardiographic findings, despite almost complete resolution of perfusion defects on lung scan. CONCLUSIONS: The extent of right ventricular dysfunction in hemodynamically stable, previously normal patients with acute pulmonary embolism does not reflect the extent of the perfusion abnormalities. Further, right ventricular enlargement and systolic dysfunction are present and persistent despite treatment with heparin and warfarin therapy or vena caval interruption.

Relation between patent foramen ovale and perfusion abnormalities in acute pulmonary embolism.

The frequency of right-to-left shunt through a patent foramen ovale is increased in hemodynamically stable patients without preexisting cardiopulmonary disease with acute pulmonary embolism. This is associated with a greater degree of perfusion abnormalities as quantified by perfusion scan and the presence of tricuspid regurgitation.

Thromboxane-mediated alveolar responses to acute obstruction of the pulmonary vasculature.

To investigate the mechanisms responsible for the accelerated radioaerosol transalveolar clearance rates caused by acute balloon occlusion of the pulmonary arterial tree in dogs, the effects of vagal tone (n = 5) and prostaglandins (n = 21) including thromboxane were evaluated. In 15 animals, serial pulmonary arterial and aortic blood samples were acquired to evaluate pertinent metabolic products. Balloon occlusion of a localized arterial territory caused significant acceleration of technetium-99m diethylenetriamine pentaacetic acid (Tc-99m DTPA) clearance in the zone immediately distal to the occlusion (baseline clearance half-time 23.1 +/- 0/7 min versus 19.3 +/- 0.4 min, mean +/- SEM, p < 0.05). Vagotomy had no effect on occlusion-accelerated clearance. However, significant (p < 0.05) normalization did occur in the presence of indomethacin (21.9 +/- 0.4 min) and meclofenamic acid (20.4 +/- 0.5 min). Plasma values of thromboxane rose dramatically (pulmonary blood baseline 119 pg/ml to > 40,000 pg/ml) and transiently immediately after pulmonary vascular occlusion, and this rise was blunted significantly (peak pulmonary thromboxane B2 [TXB2] concentration = 668 pg/ml, p < 0.05) by meclofenamic acid. Significant normalization of local DTPA clearance rates also occurred in the presence of a thromboxane receptor blocker (n = 4), even when blood levels of thromboxane were elevated. Changes in transalveolar DTPA clearance rates after balloon occlusion of pulmonary arteries seem to a significant extent to be thromboxane-mediated.

Effect of pulmonary volume loss on the size and appearance of experimentally produced scintigraphic perfusion defects.

RATIONALE AND OBJECTIVES: Atelectasis is a frequent concomitant of pulmonary embolism (PE) and other conditions that mimic PE. Accordingly, we developed an animal model to study the effect of regional loss of pulmonary volume on the size and configuration of experimentally induced perfusion defects that simulated PE. METHODS: After baseline anteroposterior and lateral chest radiographs were taken, 11 pulmonary vascular occlusions were created in 10 anesthetized, intubated mongrel dogs. Three were created by balloon occlusions of segmental pulmonary arteries, and eight were created by release of autologous PE. The size and shape of the affected lung zones were determined by perfusion scintigraphy (technetium-99m-macroaggregrated albumin [99mTc-MAA]) and the results recorded. A balloon-tip Fogarty catheter then was passed through the trachea and inflated at preselected endobronchial sites to produce bronchial occlusion and volume loss, which were documented radiographically. Because one animal had two such occlusions sequentially, a total of 12 volume reduction experiments were performed. After inducing volume loss, perfusion imaging was repeated without reinjection of 99mTc-MAA to allow any changes in the perfusion defects to be recorded, measured, and compared with the degree of volume loss visualized radiographically. RESULTS: Volume loss did not alter the shape of ipsilateral scintigraphic perfusion defects, but it did result in the expansion of eight of the 12 perfusion defects (range = 18-137% increase in area, Mdn = 33%) and a 28% decrease in the size of one defect located within a lobe that lost volume. Despite these quantitative changes, only three of the perfusion defects appeared substantially larger. When independently interpreted radiographs demonstrated large volume losses, quantitative changes in perfusion defects were no greater on average than when radiographic volume loss was modest. CONCLUSION: Loss of pulmonary volume is occasionally associated with considerable ipsilateral alterations in the size of scintigraphic perfusion defects. It is probably advisable to consider the effects of anatomic distortions such as volume loss when interpreting the significance of scintigraphic perfusion defects in patients suspected of having PE.

Unlocking clinical data from narrative reports: a study of natural language processing.

OBJECTIVE: To evaluate the automated detection of clinical conditions described in narrative reports. DESIGN: Automated methods and human experts detected the presence or absence of six clinical conditions in 200 admission chest radiograph reports. STUDY SUBJECTS: A computerized, general-purpose natural language processor; 6 internists; 6 radiologists; 6 lay persons; and 3 other computer methods. MAIN OUTCOME MEASURES: Intersubject disagreement was quantified by "distance" (the average number of clinical conditions per report on which two subjects disagreed) and by sensitivity and specificity with respect to the physicians. RESULTS: Using a majority vote, physicians detected 101 conditions in the 200 reports (0.51 per report); the most common condition was acute bacterial pneumonia (prevalence, 0.14), and the least common was chronic obstructive pulmonary disease (prevalence, 0.03). Pairs of physicians disagreed on the presence of at least 1 condition for an average of 20% of reports. The average intersubject distance among physicians was 0.24 (95% Cl, 0.19 to 0.29) out of a maximum possible distance of 6. No physician had a significantly greater distance than the average. The average distance of the natural language processor from the physicians was 0.26 (Cl, 0.21 to 0.32; not significantly greater than the average among physicians). Lay persons and alternative computer methods had significantly greater distance from the physicians (all > 0.5). The natural language processor had a sensitivity of 81% (Cl, 73% to 87%) and a specificity of 98% (Cl, 97% to 99%); physicians had an average sensitivity of 85% and an average specificity of 98%. CONCLUSIONS: Physicians disagreed on the interpretation of narrative reports, but this was not caused by outlier physicians or a consistent difference in the way internists and radiologists read reports. The natural language processor was not distinguishable from the physicians and was superior to all other comparison subjects. Although the domain of this study was restricted (six clinical conditions in chest radiographs), natural language processing seems to have the potential to extract clinical information from narrative reports in a manner that will support automated decision-support and clinical research.

Effect of acute experimental pulmonary arterial occlusion on the deposition and clearance of technetium-99m-DTPA radioaerosols.

UNLABELLED: Canine models of autologous PE and balloon occlusion of the pulmonary vasculature were used to evaluate radioaerosol deposition and to quantify radioaerosol clearance in the immediate postembolic period. METHODS: A total of 28 animals were anesthetized, intubated and studied (central balloon occlusion = 4, peripheral balloon occlusion = 7, autologous PE = 5, various control groups = 12). A gamma camera computer system was used to monitor the distribution and clearance rate of inhaled DTPA radioaerosol. The perfusion defect distribution was determined after the radioaerosol study using 99mTc-MAA. RESULTS: A new radioaerosol deposition defect was seen in 3 of 16 animals in a zone of acute vascular occlusion by a balloon (n = 2) or PE (n = 1). In addition, radioaerosol clearance rates were altered substantially. Peripheral vascular occlusion and PE caused radioaerosol clearance rates to accelerate significantly (control clearance half-time = 33.6 +/- 4.6 min, post-balloon occlusion = 19.0 +/- 8.3 min, post-PE = 12.4 +/_ 3.7 min, both p < 0.05). In only one case did these clearance rate changes create a visible abnormality in the aerosol images. CONCLUSIONS: The pulmonary deposition patterns and clearance rates of 99mTc-DTPA radioaerosol can be altered by acute vascular occlusion or PE. These findings should be considered when interpreting radioaerosol images in patients with suspected PE.

A general natural-language text processor for clinical radiology.

OBJECTIVE: Development of a general natural-language processor that identifies clinical information in narrative reports and maps that information into a structured representation containing clinical terms. DESIGN: The natural-language processor provides three phases of processing, all of which are driven by different knowledge sources. The first phase performs the parsing. It identifies the structure of the text through use of a grammar that defines semantic patterns and a target form. The second phase, regularization, standardizes the terms in the initial target structure via a compositional mapping of multi-word phrases. The third phase, encoding, maps the terms to a controlled vocabulary. Radiology is the test domain for the processor and the target structure is a formal model for representing clinical information in that domain. MEASUREMENTS: The impression sections of 230 radiology reports were encoded by the processor. Results of an automated query of the resultant database for the occurrences of four diseases were compared with the analysis of a panel of three physicians to determine recall and precision. RESULTS: Without training specific to the four diseases, recall and precision of the system (combined effect of the processor and query generator) were 70% and 87%. Training of the query component increased recall to 85% without changing precision.