Dataset Information

Automatic Determination of the Need for Intravenous Contrast in Musculoskeletal MRI Examinations Using IBM Watson's Natural Language Processing Algorithm.

ABSTRACT: Magnetic resonance imaging (MRI) protocoling can be time- and resource-intensive, and protocols can often be suboptimal dependent upon the expertise or preferences of the protocoling radiologist. Providing a best-practice recommendation for an MRI protocol has the potential to improve efficiency and decrease the likelihood of a suboptimal or erroneous study. The goal of this study was to develop and validate a machine learning-based natural language classifier that can automatically assign the use of intravenous contrast for musculoskeletal MRI protocols based upon the free-text clinical indication of the study, thereby improving efficiency of the protocoling radiologist and potentially decreasing errors. We utilized a deep learning-based natural language classification system from IBM Watson, a question-answering supercomputer that gained fame after challenging the best human players on Jeopardy! in 2011. We compared this solution to a series of traditional machine learning-based natural language processing techniques that utilize a term-document frequency matrix. Each classifier was trained with 1240 MRI protocols plus their respective clinical indications and validated with a test set of 280. Ground truth of contrast assignment was obtained from the clinical record. For evaluation of inter-reader agreement, a blinded second reader radiologist analyzed all cases and determined contrast assignment based on only the free-text clinical indication. In the test set, Watson demonstrated overall accuracy of 83.2% when compared to the original protocol. This was similar to the overall accuracy of 80.2% achieved by an ensemble of eight traditional machine learning algorithms based on a term-document matrix. When compared to the second reader's contrast assignment, Watson achieved 88.6% agreement. When evaluating only the subset of cases where the original protocol and second reader were concordant (n = 251), agreement climbed further to 90.0%. The classifier was relatively robust to spelling and grammatical errors, which were frequent. Implementation of this automated MR contrast determination system as a clinical decision support tool may save considerable time and effort of the radiologist while potentially decreasing error rates, and require no change in order entry or workflow.

SUBMITTER: Trivedi H

PROVIDER: S-EPMC5873465 | biostudies-literature | 2018 Apr

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Automatic Determination of the Need for Intravenous Contrast in Musculoskeletal MRI Examinations Using IBM Watson's Natural Language Processing Algorithm.

Trivedi Hari H Mesterhazy Joseph J Laguna Benjamin B Vu Thienkhai T Sohn Jae Ho JH

Journal of digital imaging 20180401 2

Magnetic resonance imaging (MRI) protocoling can be time- and resource-intensive, and protocols can often be suboptimal dependent upon the expertise or preferences of the protocoling radiologist. Providing a best-practice recommendation for an MRI protocol has the potential to improve efficiency and decrease the likelihood of a suboptimal or erroneous study. The goal of this study was to develop and validate a machine learning-based natural language classifier that can automatically assign the u ...[more]

PMID: 28924815

Similar Datasets

Project description:PurposeThe global noise (GN) algorithm has been previously introduced as a method for automatic noise measurement in clinical CT images. The accuracy of the GN algorithm has been assessed in abdomen CT examinations, but not in any other body part until now. This work assesses the GN algorithm accuracy in automatic noise measurement in head CT examinations.MethodsA publicly available image dataset of 99 head CT examinations was used to evaluate the accuracy of the GN algorithm in comparison to reference noise values. Reference noise values were acquired using a manual noise measurement procedure. The procedure used a consistent instruction protocol and multiple observers to mitigate the influence of intra- and interobserver variation, resulting in precise reference values. Optimal GN algorithm parameter values were determined. The GN algorithm accuracy and the corresponding statistical confidence interval were determined. The GN measurements were compared across the six different scan protocols used in this dataset. The correlation of GN to patient head size was also assessed using a linear regression model, and the CT scanner's X-ray beam quality was inferred from the model fit parameters.ResultsAcross all head CT examinations in the dataset, the range of reference noise was 2.9-10.2 HU. A precision of ±0.33 HU was achieved in the reference noise measurements. After optimization, the GN algorithm had a RMS error 0.34 HU corresponding to a percent RMS error of 6.6%. The GN algorithm had a bias of +3.9%. Statistically significant differences in GN were detected in 11 out of the 15 different pairs of scan protocols. The GN measurements were correlated with head size with a statistically significant regression slope parameter (p < 10-7 ). The CT scanner X-ray beam quality estimated from the slope parameter was 3.5 cm water HVL (2.8-4.8 cm 95% CI).ConclusionThe GN algorithm was validated for application in head CT examinations. The GN algorithm was accurate in comparison to reference manual measurement, with errors comparable to interobserver variation in manual measurement. The GN algorithm can detect noise differences in examinations performed on different scanner models or using different scan protocols. The trend in GN across patients of different head sizes closely follows that predicted by a physical model of X-ray attenuation.

Project description:BackgroundApproximately 1% of low back pain is estimated to be caused by serious systemic diseases, including cancer, infection, or abdominal aortic dissection. This study aimed to determine the frequency of execution of non-MSK physical examination procedures among Quebec chiropractors and to identify the clinical context that prompts them to use these physical examination procedures.MethodsCross-sectional survey containing 44 questions administered to a random sample of Quebec chiropractors using a succession of online, postal and phone questionnaires. The 4-part survey questionnaire contained six demographic questions, 28 single-choice questions to determine the frequency of execution of non-MSK physical examination procedures, seven short clinical vignettes for which the respondents had to select the non-MSK examinations that would be required, and two questions inquiring about the proportion of new patients for which participants' felt non-MSK examinations were necessary and whether appropriate assessments were performed. The questionnaire was pilot tested, and feedback received integrated prior to administration. We conducted descriptive statistics, Pearson correlations, and an ANOVA.ResultsThe survey was completed by 182 chiropractors (response rate: 36.4%). The most commonly non-musculoskeletal examination performed daily were blood pressure (12.1%) and cranial nerves (4.9%). The most common tests never performed were oxygen saturation (68.7%), cardiac auscultation (69.2%), tibio-brachial index (71.4%), breast (86.8%), rectal (96.7%), testicular (95.6%), and vaginal (99.9%) exams. Female chiropractors and Quebec University in Trois-Rivières graduates reported that a significantly higher proportion of their new patients required a non-musculoskeletal physical examination compared to male participants (37.2% vs 28.3%) or Canadian Memorial Chiropractic College graduates (33.9% vs 19.9%). Reason for not performing a physical examination included the belief that another healthcare professional was better positioned to perform and/or interpret the related tests (76.4%).ConclusionsVital signs and cranial nerve examinations were the most frequency performed non-musculoskeletal examinations reported by chiropractors. Apart from the genitourinary exam almost never performed, most participants chose non-musculoskeletal examinations deemed appropriate for the patient's presentation.

Project description:To determine whether intravenous low-osmolality iodinated contrast material is associated with post-computed tomography (CT) acute kidney injury (AKI).Institutional review board approval was obtained and patient consent waived for this HIPAA-compliant retrospective study. CT examinations performed over a 10-year period in adult inpatients with sufficient serum creatinine (SCr) data were identified. A one-to-one propensity-matched matched cohort analysis with multivariate analysis of effects was performed with post-CT AKI as the primary outcome measure (10,121 unenhanced and 10,121 intravenous contrast-enhanced CT examinations in 20,242 patients). Propensity matching was performed with respect to likelihood of patient receiving intravenous contrast material (36 tested covariates). The primary endpoint was post-CT AKI by using Acute Kidney Injury Network SCr criteria; the secondary endpoint was post-CT AKI by using traditional SCr criteria for contrast material-induced nephrotoxicity (CIN; SCr increase ≥0.5 mg/dL [44.20 μmol/L] or ≥25%). Multivariate subgroup threshold analysis was performed (SCr <1.5 [<132.60 μmol/L]; ≥1.5 to ≥2.0 mg/dL [≥132.60 to ≥176.80 μmol/L]) and adjusted for assigned propensity scores.Intravenous low-osmolality iodinated contrast material had a significant effect on the development of post-CT AKI for patients with pre-CT SCr levels of 1.6 mg/dL (141.44 μmol/L) or greater (odds ratio, 1.45; 95% confidence interval [CI]: 1.11, 1.89;P = .007). This effect strengthened as pre-CT SCr increased. Patients with stable SCr less than 1.5 mg/dL (132.60 μmol/L) were not at risk for developing CIN (P = .25, power > 95%). Both endpoints demonstrated similar results (eg, SCr ≥1.6 mg/dL [141.44 μmol/L] by using traditional CIN criteria: odds ratio, 1.64; 95% CI: 1.18, 2.28; P = .003). Post-CT AKI was prevalent in both the unenhanced and contrast-enhanced CT subgroups, and it increased with increases in pre-CT SCr. Many risk factors contributed to development of post-CT AKI, regardless of iodinated contrast material.Intravenous low-osmolality iodinated contrast material is a nephrotoxic risk factor, but not in patients with a stable SCr level less than 1.5 mg/dL. Many factors other than contrast material can affect post-CT AKI rates.

Project description:To propose a method to quantify T1 and contrast agent uptake in breast dynamic contrast-enhanced (DCE) examinations undertaken with standard clinical fat-suppressed MRI sequences and to demonstrate the proposed approach by comparing the enhancement characteristics of lobular and ductal carcinomas.A standard fat-suppressed DCE of the breast was performed at 1.5 T (Siemens Aera), followed by the acquisition of a proton density (PD)-weighted sequence, also fat suppressed. Both sequences were characterized with test objects (T1 ranging from 30 ms to 2,400 ms) and calibration curves were obtained to enable T1 calculation. The reproducibility and accuracy of the calibration curves were also investigated. Healthy volunteers and patients were scanned with Ethics Committee approval. The effect of B0 field inhomogeneity was assessed in test objects and healthy volunteers. The T1 of breast tumors was calculated at different time points (pre-, peak-, and post-contrast agent administration) for 20 patients, pre-treatment (10 lobular and 10 ductal carcinomas) and the two cancer types were compared (Wilcoxon rank-sum test).The calibration curves proved to be highly reproducible (coefficient of variation under 10%). T1 measurements were affected by B0 field inhomogeneity, but frequency shifts below 50 Hz introduced only 3% change to fat-suppressed T1 measurements of breast parenchyma in volunteers. The values of T1 measured pre-, peak-, and post-contrast agent administration demonstrated that the dynamic range of the DCE sequence was correct, that is, image intensity is approximately directly proportional to 1/T1 for that range. Significant differences were identified in the width of the distributions of the post-contrast T1 values between lobular and ductal carcinomas (P < 0.05); lobular carcinomas demonstrated a wider range of post-contrast T1 values, potentially related to their infiltrative growth pattern.This work has demonstrated the feasibility of fat-suppressed T1 measurements as a tool for clinical studies. The proposed quantitative approach is practical, enabled the detection of differences between lobular and invasive ductal carcinomas, and further enables the optimization of DCE protocols by tailoring the dynamic range of the sequence to the values of T1 measured.

Dataset Information

Automatic Determination of the Need for Intravenous Contrast in Musculoskeletal MRI Examinations Using IBM Watson's Natural Language Processing Algorithm.

Publications

Automatic Determination of the Need for Intravenous Contrast in Musculoskeletal MRI Examinations Using IBM Watson's Natural Language Processing Algorithm.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets