Extracting hypertrophic cardiomyopathy features from cardiac magnetic resonance reports by natural language processing

Objective

Determine if information regarding hypertrophic cardiomyopathy (HCM) can be accurately retrieved from cardiac magnetic resonance (CMR) reports using natural language processing (NLP).

Background

CMR imaging is used for diagnosis and risk stratification of HCM. Manual annotation of information from CMR is time-consuming. NLP is an artificial intelligence method for automating extraction of information from narrative text.

Methods

We identified 200 HCM patients who had CMR reports from 1998 to 2018. These patients were randomly allocated into training (100 patients with 185 CMR reports) and testing sets (100 patients with 206 reports). An NLP system with 2 tiers was developed; the first extracted information regarding HCM diagnosis while second extracted categorical or numeric concepts for HCM classification. NLP performance was compared with gold-standard manual annotation.

Results

NLP algorithms achieved very high performance across all concepts with mean positive predictive value (PPV) = 0.96. An outlier was the performance for abstracting the presence of an apical pouch from CMR reports, which had noticeably lower PPV= 0.78 which be attributed to the low number of cases with this finding.

Conclusions

The algorithms developed can be translated to clinical decision support systems to increase efficiency and contribute to improved quality of care.

Funding Acknowledgement

Type of funding source: Other. Main funding source(s): Study supported by the National Heart, Lung and Blood Institute of National Institutes of Health (K01HL124045), the Mayo Clinic Center for Clinical and Translational Science (CCaTS), and the Mayo Clinic K2R award. Content is solely the responsibility of authors and does not necessarily represent official views of the National Institutes of Health.

Multisite Concordance of DSC-MRI Analysis for Brain Tumors: Results of a National Cancer Institute Quantitative Imaging Network Collaborative Project

BACKGROUND AND PURPOSE

Standard assessment criteria for brain tumors that only include anatomic imaging continue to be insufficient. While numerous studies have demonstrated the value of DSC-MR imaging perfusion metrics for this purpose, they have not been incorporated due to a lack of confidence in the consistency of DSC-MR imaging metrics across sites and platforms. This study addresses this limitation with a comparison of multisite/multiplatform analyses of shared DSC-MR imaging datasets of patients with brain tumors.

MATERIALS AND METHODS

DSC-MR imaging data were collected after a preload and during a bolus injection of gadolinium contrast agent using a gradient recalled-echo-EPI sequence (TE/TR = 30/1200 ms; flip angle = 72°). Forty-nine low-grade (n = 13) and high-grade (n = 36) glioma datasets were uploaded to The Cancer Imaging Archive. Datasets included a predetermined arterial input function, enhancing tumor ROIs, and ROIs necessary to create normalized relative CBV and CBF maps. Seven sites computed 20 different perfusion metrics. Pair-wise agreement among sites was assessed with the Lin concordance correlation coefficient. Distinction of low- from high-grade tumors was evaluated with the Wilcoxon rank sum test followed by receiver operating characteristic analysis to identify the optimal thresholds based on sensitivity and specificity.

RESULTS

For normalized relative CBV and normalized CBF, 93% and 94% of entries showed good or excellent cross-site agreement (0.8 ≤ Lin concordance correlation coefficient ≤ 1.0). All metrics could distinguish low- from high-grade tumors. Optimum thresholds were determined for pooled data (normalized relative CBV = 1.4, sensitivity/specificity = 90%:77%; normalized CBF = 1.58, sensitivity/specificity = 86%:77%).

CONCLUSIONS

By means of DSC-MR imaging data obtained after a preload of contrast agent, substantial consistency resulted across sites for brain tumor perfusion metrics with a common threshold discoverable for distinguishing low- from high-grade tumors.

The relation of hedonic hunger and restrained eating to lateralized frontal activation

Asymmetrical alpha activation in the prefrontal cortex (frontal asymmetry) in electroencephalography (EEG) has been related to eating behavior. Prior studies linked dietary restraint with right frontal asymmetry [1] and disinhibition with left frontal asymmetry [2]. The current study simultaneously assessed restrained eating and hedonic hunger (drive for food reward in the absence of hunger) in relation to frontal asymmetry. Resting-state EEG and measures of restrained eating (Revised Restraint Scale; RRS) and hedonic hunger (Power of Food Scale; PFS) were assessed in 61 non-obese adults. Individually, hedonic hunger predicted left asymmetry. However, PFS and RRS were correlated (r=0.48, p<0.05) and there was a significant interaction between PFS and RRS on frontal asymmetry, p<0.01. Results indicated that those high in hedonic hunger exhibited left asymmetry irrespective of RRS scores; among those low in PFS, only those high in RRS showed right asymmetry. Results were consistent with literature linking avoidant behaviors (restraint) with right-frontal asymmetry and approach behaviors (binge eating) with left-frontal asymmetry. It appears that a strong drive toward palatable foods predominates at a neural level even when restraint is high. Findings suggest that lateralized frontal activity is an indicator of motivation both to consume and to avoid consuming highly palatable foods.

SU-E-T-252: Consolidating Duodenal/Small Bowel Toxicity Data via Iso-Effective Dose Calculations Based on Compiled Clinical Data

PURPOSE

To consolidate duodenal toxicity data from clinical studies with different dose fractionation schemes using the modified linear quadratic (MLQ) model. A methodology of adjusting the dose-volume parameters todifferent levels of normal tissue complication probability (NTCP) was proposed and used to estimate dose-volume constrains for treatment planning.

METHODS

A set of modified Lyman model parameters for duodenum NTCP were estimated by the chi-squared fitting method using tolerance dose and equivalent uniform dose (EUD) data obtained in a literature search. These model parameters were then used to convert the dose-volume pair, (D, V) to the iso-effective dose (in 2 Gy per fraction)- volume pair, (DMLQED2, V). A relationship was derived to convert a given DMLQED2 at one level of NTCP, to an iso-effective dose at another NTCP.

RESULTS

The literature search yielded six reports useful in making estimates of small bowel/duodenal toxicity. The modified Lyman model parameters were found to be TD50 = 60.9 ± 7.9 Gy, m = 0.21 ± 0.05, and Î = 0.09 ± 0.03 Gy-1. The toxicity rates associated with hypo-fractionated radiotherapy (HBRT) were found to be consistent with other clinical data of conventional fractionations found in the literature. The conversion of DMLQED2 between different NTCP levels remains consistent with each other over a narrow range of NTCP.

CONCLUSION

MLQ based iso-effective calculations of dose-response data corresponding to Grade > 2 toxicity were found to be consistent with one another within the uncertainty of DMLQED2 due to model parameter uncertainty. The dose-volume data that can be converted to different NTCP levels may be used to estimate duodenal/small bowel dose-volume constrains for new dose fractionation and/or dose escalation strategies. Medical College of Wisconsin Cancer Center Meinerz Foundation.

Diffusion-weighted imaging characteristics of biopsy-proven demyelinating brain lesions

OBJECTIVES

To describe lesional diffusion-weighted imaging characteristics in a cohort of patients with biopsy-proven CNS inflammatory demyelinating disease (IDD) and compare diffusion characteristics of ring-enhancing CNS IDD lesions vs abscesses and tumors.

METHODS

Forty prebiopsy apparent diffusion coefficient (ADC) maps were reviewed from 30 patients with CNS IDD. Lesions were analyzed for size, T2-weighted (T2W) hypointense rim, enhancement, and ADC pattern. ADC patterns of CNS IDD ring-enhancing lesions were compared with a published cohort of 35 patients with ring-enhancing tumors and abscesses.

RESULTS

IDD lesions displayed a spectrum of peripheral ADC patterns at the lesion edge: restricted diffusion (low ADC), 33%; increased diffusion (high ADC), 60%; and normal diffusion (homogeneously isointense), 7%. Of biopsied lesions, 93% enhanced (ring, 52%; heterogeneous, 34%; homogeneous, 7%). A hypointense T2W rim was observed in 53%. A ring pattern on ADC (isointense or dark) was associated with T2W hypointense rims (p = 0.02) but not with ring enhancement. On serial imaging, 4 of 7 (57%) patients demonstrated changes in ADC patterns. Peripheral restriction was more common in IDD (p = 0.006) than in tumors or abscesses, whereas central restriction was only observed in abscesses. Restricted lesions in the same stage were more common in the non-IDD cohort (42% vs 20%), with a uniform restricted pattern seen only in abscesses.

CONCLUSIONS

In ring-enhancing lesions, peripheral diffusion restriction is more common in IDD than in tumors/abscesses, whereas central restriction is more common among abscesses. Rapid ADC pattern changes in IDD probably reflect dynamic lesion evolution and may distinguish IDD from tumors.

Visualization of 3D osteon morphology by synchrotron radiation micro-CT

Cortical bone histology has been the subject of scientific inquiry since the advent of the earliest microscopes. Histology - literally the study of tissue - is a field nearly synonymous with 2D thin sections. That said, progressive developments in high-resolution X-ray imaging are enabling 3D visualization to reach ever smaller structures. Micro-computed tomography (micro-CT), employing conventional X-ray sources, has become the gold standard for 3D analysis of trabecular bone and is capable of detecting the structure of vascular (osteonal) porosity in cortical bone. To date, however, direct 3D visualization of secondary osteons has eluded micro-CT based upon absorption-derived contrast. Synchrotron radiation micro-CT, through greater image quality, resolution and alternative contrast mechanisms (e.g. phase contrast), holds great potential for non-destructive 3D visualization of secondary osteons. Our objective was to demonstrate this potential and to discuss areas of bone research that can be advanced through the application of this approach. We imaged human mid-femoral cortical bone specimens derived from a 20-year-old male (Melbourne Femur Collection) at the Advanced Photon Source synchrotron (Chicago, IL, USA) using the 2BM beam line. A 60-mm distance between the target and the detector was employed to enhance visualization of internal structures through propagation phase contrast. Scan times were 1 h and images were acquired with 1.4-μm nominal isotropic resolution. Computer-aided manual segmentation and volumetric 3D rendering were employed to visualize secondary osteons and porous structures, respectively. Osteonal borders were evident via two contrast mechanisms. First, relatively new (hypomineralized) osteons were evident due to differences in X-ray attenuation relative to the surrounding bone. Second, osteon boundaries (cement lines) were delineated by phase contrast. Phase contrast also enabled the detection of soft tissue remnants within the vascular pores. The ability to discern osteon boundaries in conjunction with vascular and cellular porosity revealed a number of secondary osteon morphologies and provided a unique 3D perspective of the superimposition of secondary osteons on existing structures. Improvements in resolution and optimization of the propagation phase contrast promise to provide further improvements in structural detail in the future.

Helical tomotherapy for anal cancer: Initial clinical outcomes and organ motion

530

Background: Helical tomotherapy (HT) delivers rotational IMRT with megavoltage CT (MVCT) guidance for patient positioning, offering improved dose conformality and accuracy. We report preliminary clinical outcomes using HT in patients with anal cancer with estimated organ motion and delivered doses.

Methods: From 9/2005-8/2009, 15 patients with anal cancer were treated with HT and concurrent chemotherapy (MMC 10 mg/m2 IV bolus d1 and 29; 5FU 1,000 mg/m2/day CI d1-4 and 29-32). Median doses for primary tumor, involved lymph nodes and uninvolved lymph nodes were 50.5, 50.4 and 45 Gy. Dose specifications and target definitions were per RTOG 0529 for the majority. Treatment plans, daily MVCT, and initial clinical outcomes and toxicity were retrospectively analyzed. Center-of-mass (COM) locations and volumes of the targets and OAR with daily shifts were analyzed on selected patients. Doses actually delivered daily were reconstructed.

Results: At median follow-up of 15 months, all patients were alive without any evidence of recurrence. There were 40% male, 27% HIV+, 13% T3/4, and 47% lymph node positive. Four patients required treatment break of ≤ 1 week. There were no non-hematologic grade 4/5 toxicities. Overall non-hematologic grade 3 toxicity was 33.3% (GI and skin). All other patients had ≤ grade 2 GI or skin toxicity that completely resolved with aggressive supportive care. There was 40% overall hematologic grade 4 toxicity. Overall hematologic grade 3 toxicity was 60%. Daily shifts were 0.6 ± 3.3, 0.4 ± 3.0, and 0.5 ± 2.2 (SD) mm in lateral, longitudinal and vertical directions. Average daily displacements of the COM from their locations on the planning CT were 9 ± 3, 7 ± 4, 9 ± 2, and 7 ± 2 (SD) mm, for anus, rectum, bladder and vulva. Interfractional organ motion and deformation resulted in variations in the dose delivered in each fraction.For example, daily dose covering 95% of PTV varied by 4%, and vulva volume covered by 50 Gy changed by 23% daily.

Conclusions: HT with concurrent chemotherapy is well-tolerated compared to historical reports. Interfractional organ motion is significant and can result in deviations in dose delivery to the target and OAR. Such deviations should be considered in analyzing dose response outcome.

No significant financial relationships to disclose.