A Pilot Study of Contextual UMLS Indexing to Improve the Precision of Concept-based Representation in XML-structured Clinical Radiology Reports

Background: Severe WMD is associated with post-stroke mortality, recurrent stroke risk, cognitive decline, and poor functional outcomes. We have previously reported that “severe” WMD grade is likely when the clinical radiology report includes the words “severe, extensive, advanced, or diffuse”. We examined whether terms consistent with the Fazekas WMD grading scale (none, mild, moderate, or severe) in clinical radiology reports can be used to categorize WMD severity in epidemiologic stroke studies, and sought to repeat our factor analysis with descriptive terms. Methods: Clinical reports from 688 ischemic stroke subjects with MRI or CT films from 2010 were reviewed by physician investigators, who recorded whether WMD was described and whether the Fazekas terms or similar descriptors (such as diffuse, patchy, extensive, etc.) were used. A stroke neurologist and a research assistant independently evaluated available neuroimaging studies and categorized WMD severity according to the Fazekas grades. WMD was preferentially graded on MRI scans; CT scans were assessed if MRIs were unavailable or not performed. Kappa statistic was used to compare the grade mentioned in the report with our direct review; factor analysis was applied to the descriptor terms and logistic regression performed to examine predictive value of descriptor terms with WMD grade from direct review. Results: Of the 688, 276 had WMD radiologist grades available, 222 had no grades and no descriptors, and 190 had no grades but had descriptors. For all films with grades available, the weighted kappa score was 0.30, indicating poor agreement between the radiologist’s WMD grade and the reviewer’s WMD grade. Examining only the 231 MRI studies with WMD grades did not improve the weighted kappa score (0.34). Factor analysis found clusters of descriptors that were significantly associated with WMD grades: “scattered, minimal, tiny, punctate, spotty” with mild (p=0.0001); “multiple, patchy, diffuse” with moderate (p=0.01); and “advanced, confluent, extensive” with severe (p = 0.002). Discussion: Fazekas terms in clinical radiology reports do not seem to be useful by themselves, but descriptors used in clinical radiology reports may be utilized to approximate PVWMD severity.

Download Full-text

Performance of an Artificial Intelligence-Based Platform Against Clinical Radiology Reports for the Evaluation of Noncontrast Chest CT

Academic Radiology ◽

10.1016/j.acra.2021.02.007 ◽

2021 ◽

Author(s):

Basel Yacoub ◽

Ismail M. Kabakus ◽

U. Joseph Schoepf ◽

Vincent M. Giovagnoli ◽

Andreas M. Fischer ◽

...

Keyword(s):

Artificial Intelligence ◽

Chest Ct ◽

Radiology Reports ◽

Clinical Radiology

Download Full-text

Patients presenting with acute coronary syndromes have unreported coronary artery calcium on historical CT imaging

European Heart Journal ◽

10.1093/eurheartj/ehab724.2528 ◽

2021 ◽

Vol 42 (Supplement_1) ◽

Author(s):

C Bradley ◽

A Aggarwal ◽

K Goatman ◽

G Jones ◽

C Berry ◽

...

Keyword(s):

Risk Factors ◽

Risk Assessment ◽

Coronary Artery ◽

Cardiovascular Events ◽

Electronic Patient Record ◽

Coronary Calcification ◽

Ct Imaging ◽

Radiology Reports ◽

Clinical Radiology ◽

The Mean

Abstract Introduction Ischaemic heart disease (IHD) remains the leading cause of mortality globally1. The presence and extent of coronary artery calcification (CAC) is a strong predictor of cardiovascular events, and CAC scoring has been shown to be more predictive of cardiovascular events than other traditional risk assessment scores2. Incidental coronary calcification can be detected and quantified on non-gated CT chest scans covering the heart in the field of view3. This finding is typically not reported4 and hence an opportunity to optimise cardiovascular risk assessment and treatment is missed. Purpose We sought to investigate whether patients presenting to our centre with an acute coronary syndrome (ACS) event had historical CT imaging demonstrating coronary artery calcification. Methods We retrospectively reviewed case records for all patients referred to our centre for an invasive coronary angiogram following their first known admission with an ACS event. ACS were defined according to contemporary guidelines from the European Society of Cardiology. We reviewed a 3 month period prior to the COVID-19 pandemic (01/01/2019–31/03/2019). The national imaging database was interrogated to identify previous CT imaging that includes the heart in the field of view. The presence of coronary calcification was confirmed and quantified using an ordinal scoring method previously described3. The clinical radiology reports for the scans were reviewed to determine the frequency of CAC being reported. Demographic information was collected from our electronic patient record including the presence of risk factors for IHD. Prescribed medication prior to admission was also recorded using the on-admission medicines reconciliation documented in the electronic patient record. Results 385 patients with first presentation of ACS were identified. 75 (19%) had a prior non-gated CT chest imaging. The most common indication for CT was for investigation of possible malignancy. The mean interval from CT imaging to ACS admission was 36 months. CAC was present on 67 (89%) scans. The mean ordinal score was 4.04, corresponding to moderate CAC. The distribution of CAC by coronary artery revealed the majority of disease to involve the left anterior descending artery (Table 1). Only 12/67 (18%) of clinical radiology reports mentioned coronary calcification (Figure 1). Patients with CAC frequently had additional risk factors for IHD. Despite this only 42% were prescribed antiplatelet therapy, and only 45% prescribed a statin. Conclusions A significant proportion of ACS admissions have evidence of CAC on historical CT scans. This finding is often not reported and the majority of patients with demonstrated coronary artery disease are not prescribed appropriate preventative therapies. Systematic reporting of this finding may have a significant impact on the prevention of acute cardiovascular events. FUNDunding Acknowledgement Type of funding sources: None. Table 1

Download Full-text

Consistent tumor measurement reporting in serial CT scans: A pilot study.

Journal of Clinical Oncology ◽

10.1200/jco.2014.32.30_suppl.99 ◽

2014 ◽

Vol 32 (30_suppl) ◽

pp. 99-99

Author(s):

Terri P. Wolf ◽

Dana Ann Little ◽

Scott Christensen ◽

Natasha Perkins

Keyword(s):

Quality Improvement ◽

Pilot Study ◽

Response To Therapy ◽

Ct Scans ◽

Medical Provider ◽

Cancer Centers ◽

Radiology Reports ◽

Report Quality ◽

Serial Ct ◽

Measurement Consistency

99 Background: Oncologists in community cancer centers affiliated with the UC Davis Cancer Care Network reported quality concerns with inconsistently measured lesions on serial Computed Tomography (CT) scan reports. Radiologic imaging is an important tool in diagnosis, staging, and assessment of response to therapy in cancer treatment cancer. Accurate assessment of oncologic therapeutic efficacy is dependent on reliable radiology and report quality. The primary objective of this pilot study was to validate oncologist concerns about tumor measurement consistency in radiology reports through an audit. The secondary objective was to identify areas for quality improvement and establish process and time requirements for auditing. Two of four affiliated cancer centers (CC1 and CC2) were selected for a pilot assessment. Methods: CC1 and CC2 identified charts with CT scans in the audit timeframes. Auditors reviewed charts for serial CT scans and measureable disease in the audit date range and created a list of eligible charts. Auditors randomized the list and a sample of 61 (CC1) and 66 (CC2) charts were audited. Auditors reviewed CT orders, radiology reports for consistent lesion measurement, comparison to previous scan, addenda, service dates, radiology facilities, reading radiologist, and medical provider placing order. Results: This audit demonstrated deficiencies with lesion measurement consistency at both CC1 and CC2. At CC1 64% of radiology reports were consistently measured and at CC2 57% were consistently measured. Both CC1 and CC2 had identifiable areas for process and quality improvement. The audit required 86 hours at CC1 and 83 hours at CC2. The audit identified deficiencies in order clarity and information flow from the cancer center to the reading radiologist. At CC1 a radiologist reviewed the audited charts and reported an 85% concurrence with the auditors. A literature search did not provide benchmarks for measurement consistency in radiology reports. Conclusions: Oncologists complaints were verified regarding lesion measurement inconsistencies. This pilot demonstrated the need for quality oversight and implementation of standardized measurement (RECIST) criteria and tumor logs in the non-clinical trial environment.

Download Full-text

A quantitative approach for measuring laterality in clinical fMRI for preoperative language mapping

Neuroradiology ◽

10.1007/s00234-021-02685-z ◽

2021 ◽

Author(s):

Maria Olaru ◽

Ryan M. Nillo ◽

Pratik Mukherjee ◽

Leo P. Sugrue

Keyword(s):

Meta Analysis ◽

Population Level ◽

Language Mapping ◽

Activation Threshold ◽

Clinical Assessments ◽

Radiology Reports ◽

Clinical Radiology ◽

Clinical Fmri ◽

Language Laterality ◽

Language Areas

Abstract Purpose fMRI is increasingly used for presurgical language mapping, but lack of standard methodology has made it difficult to combine/compare data across institutions or determine the relative efficacy of different approaches. Here, we describe a quantitative analytic framework for determining language laterality in clinical fMRI that addresses these concerns. Methods We retrospectively analyzed fMRI data from 59 patients who underwent presurgical language mapping at our institution with identical imaging and behavioral protocols. First, we compared the efficacy of different regional masks in capturing language activations. Then, we systematically explored how laterality indices (LIs) computed from these masks vary as a function of task and activation threshold. Finally, we determined the percentile threshold that maximized the correlation between the results of our LI approach and the laterality assessments from the original clinical radiology reports. Results First, we found that a regional mask derived from a meta-analysis of the fMRI literature better captured language task activations than masks based on anatomically defined language areas. Then, we showed that an LI approach based on this functional mask and percentile thresholding of subject activation can quantify the relative ability of different language tasks to lateralize language function at the population level. Finally, we determined that the 92nd percentile of subject-level activation provides the optimal LI threshold with which to reproduce the original clinical reports. Conclusion A quantitative framework for determining language laterality that uses a functionally-derived language mask and percentile thresholding of subject activation can combine/compare results across tasks and patients and reproduce clinical assessments of language laterality.

Download Full-text