Skeletal muscle quantity and quality evaluation in heart failure: comparing thoracic versus abdominopelvic CT approaches.

Given the critical role of skeletal muscle in healthy aging, low muscle mass (myopenia) and quality (myosteatosis) can be used as predictors of poor functional and cardiometabolic outcomes. Myopenia is also a part of sarcopenia and malnutrition diagnostic criteria. However, there is limited evidence for using chest computed tomography (CT) to evaluate muscle health. We aimed to compare chest CT landmarks to the widely used L3 vertebra for single-slice skeletal muscle evaluation in patients with heart failure (HF). Patients admitted for acute decompensated HF between January 2017 and December 2018 were retrospectively analyzed. Body composition measurements were made on CT of the chest and abdomen/pelvis with or without contrast one month before discharge. Skeletal muscle index (SMI) and intermuscular adipose tissue percentage (IMAT%) were calculated at several thoracic levels (above the aortic arch, T8, and T12) and correlated to the widely used L3 level. A total of 200 patients were included, 89 (44.5%) female. The strongest correlation of thoracic SMI (for muscle quantity) and IMAT% (for muscle quality) with L3 was at the T12 level (r = 0.834, p < 0.001 and r = 0.757, p < 0.001, respectively). Cutoffs to identify low muscle mass for T12 SMI (derived from the lowest sex-stratified L3 SMI tertile) were 31.1 cm²/m² in men and 26.3 cm²/m² in women. SMI and IMAT% at T12 had excellent correlations with the widely used L3 level for muscle quantity and quality evaluation in patients with HF.

Sarcopenia and Frailty in Patients Undergoing Transcatheter Aortic Valve Replacement.

BACKGROUND: Skeletal muscle mass (SMM) plays a crucial role in risk assessment in transcatheter aortic valve replacement (TAVR) candidates, yet it remains underutilized. Traditional methods focus on weakness or performance but omit SMM. This study compared traditional and novel markers of sarcopenia and frailty in terms of their ability to predict adverse outcomes post-TAVR. METHODS: Three risk models were evaluated for the composite outcome of perioperative complications, 1-year rehospitalization, or 1-year mortality: (1) sarcopenia by combining low muscle mass (LMM) and weakness/performance assessed by hand grip strength or gait speed; (2) frailty by an Adapted Green score; and (3) frailty by the Green-SMI score incorporating LMM by multi-level opportunistic pre-TAVR thoracic CT segmentation. RESULTS: In this study we included 184 eligible patients from January to December of 2018, (96.7%) of which were balloon expandable valves. The three risk models identified 22.8% patients as sarcopenic, 63.6% as frail by the Adapted Green score, and 53.8% as frail by the Green-SMI score. There were higher rates of the composite outcome in patients with sarcopenia (54.8%) and frailty (41.9% with the Adapted Green and 50.5% with the Green-SMI score) compared to their non-sarcopenic (30.3%) and non-frail counterparts (25.4% with the Adapted Green and 18.8% with the Green-SMI score). Sarcopenia and frailty by Green-SMI, but not by the Adapted Green, were associated with higher risks of the composite outcome on multivariable adjustment (HR 2.2 [95% CI: 1.25-4.02], p=0.007 and HR 3.4 [95% CI: 1.75-6.65], p<0.001, respectively). CONCLUSIONS: The integration of pre-operative CT-based SMM to a frailty score significantly improves the prediction of adverse outcomes in patients undergoing TAVR.

Pre-test Prediction of Non-ischemic Cardiomyopathies using Time-Series EHR Data.

Clinical imaging is an important diagnostic test to diagnose non-ischemic cardiomyopathies (NICM). However, accurate interpretation of imaging studies often requires readers to review patient histories, a time consuming and tedious task. We propose to use time-series analysis to predict the most likely NICMs using longitudinal electronic health records (EHR) as a pseudo-summary of EHR records. Time-series formatted EHR data can provide temporality information important towards accurate prediction of disease. Specifically, we leverage ICD-10 codes and various recurrent neural network architectures for predictive modeling. We trained our models on a large cohort of NICM patients who underwent cardiac magnetic resonance imaging (CMR) and a smaller cohort undergoing echocardiogram. The performance of the proposed technique achieved good micro-area under the curve (0.8357), F1 score (0.5708) and precision at 3 (0.8078) across all models for cardiac magnetic resonance imaging (CMR) but only moderate performance for transthoracic echocardiogram (TTE) of 0.6938, 0.4399 and 0.5864 respectively. We show that our model has the potential to provide accurate pre-test differential diagnosis, thereby potentially reducing clerical burden on physicians.

Low Muscle Mass by Preprocedural Computed Tomography Is Associated With Worse Short-Term Outcomes in Transcatheter Aortic Valve Replacement Recipients.

Low muscle mass (LMM) is associated with worse outcomes in various clinical situations. Traditional frailty markers have been used for preoperative risk stratification in patients who underwent transcatheter aortic valve replacement (TAVR). However, preoperative imaging provides an opportunity to directly quantify skeletal muscle mass to identify patients at higher risk of procedural complications. We reviewed all TAVR recipients from January to December 2018 and included subjects with preprocedural chest computed tomography. Multi-slice automated measurements of skeletal muscle mass were made from the ninth to twelfth thoracic vertebrae and normalized by height squared to obtain skeletal muscle index (cm2/m2). LMM was defined as the lowest gender-stratified skeletal muscle index tertile. Strength testing was collected during pre-TAVR evaluation. Primary outcome was a composite of perioperative complications, 1-year rehospitalization, or 1-year mortality. In our cohort, 238 patients met inclusion criteria, and 80 (33.6%) were identified to have LMM. Patients with LMM were older with lower body mass index, decreased grip strength, lower hemoglobin A1c, and higher N-terminal pro-brain natriuretic peptide. They had greater rates of the composite outcome and 2-year all-cause mortality, which remained significant on multivariable adjustment (hazard ratio 1.71, 95% confidence interval 1.05 to 2.78, p = 0.030 and hazard ratio 2.31, 95% confidence interval 1.02 to 5.24, p = 0.045, respectively) compared with patients without LMM; there was no significant difference in 5-year all-cause mortality. In conclusion, LMM was associated with an increase in the primary composite outcome and 2-year all-cause mortality in TAVR recipients. Using automatic muscle processing software on pre-TAVR computed tomography scans may serve as an additional preoperative risk stratification tool.

Rubuspuyumaensis, a new species of Rosaceae from Taiwan.

Background: The genus Rubus L. (Rosaceae), comprising approximately 750 species and classified into 12 subgenera, is distributed worldwide and is one of the largest plant genera. In Taiwan, Rubus comprises 41 taxa, including 35 species, three varieties and three hybrids. Amongst the genus Rubus, the species, previously recorded as R.howii in Taiwan, was misidentified and this study recognised it as a new species. New information: Due to its distribution mainly in south-eastern Taiwan, we named this new species as Rubuspuyumaensis, after the local aborigine tribe Puyuma. Taxonomic descriptions and colour photographs of the new species are provided to assist in identification. R.puyumaensis is most similar to R.howii and R.refractus. They can be distinguished by the colour of young leaves, leaf shape, arrangement of florets, trichomes of inflorescences, size of sepal lobes, petal colour, types and trichomes of filaments and the length of stamens and pistils.

Combined Prognostic Impact of Low Muscle Mass and Hypoalbuminemia in Patients Hospitalized for Heart Failure: A Retrospective Cohort Study.

BACKGROUND: Sarcopenia and hypoalbuminemia have been identified as independent predictors of increased adverse outcomes, including mortality and readmissions, in hospitalized older adults with acute decompensated heart failure (ADHF). However, the impact of coexisting sarcopenia and hypoalbuminemia on morbidity and death in adults with ADHF has not yet been investigated. We aimed to investigate the combined effects of lower muscle mass (LMM) as a surrogate for sarcopenia and hypoalbuminemia on in-hospital and postdischarge outcomes of patients hospitalized for ADHF. METHODS AND RESULTS: A total of 385 patients admitted for ADHF between 2017 and 2020 at a single institution were retrospectively identified. Demographic and clinical data were collected, including serum albumin levels at admission and discharge. Skeletal muscle indices were derived from semi-automated segmentation software analysis on axial chest computed tomography at the twelfth vertebral level. Our analysis revealed that patients who had LMM with admission hypoalbuminemia experienced increased diagnoses of infection and delirium with longer hospital length of stay and more frequent discharge to a facility. Upon discharge, 27.9% of patients had higher muscle mass without discharge hypoalbuminemia (reference group), 9.7% had LMM without discharge hypoalbuminemia, 38.4% had higher muscle mass with discharge hypoalbuminemia, and 24.0% had LMM with discharge hypoalbuminemia; mortality rates were 37.6%, 51.4%, 48.9%, and 63.2%, respectively. 1- and 3-year mortality risks were highest in those with LMM and discharge hypoalbuminemia; this relationship remained significant over a median 23.6 (3.1-33.8) months follow-up time despite multivariable adjustments (hazard ratio, 2.03 [95% CI, 1.31-3.16]; P=0.002). CONCLUSIONS: Hospitalization with ADHF, LMM, and hypoalbuminemia portend heightened mortality risk.

The association of body composition and outcomes following autologous hematopoietic stem cell transplantation in patients with non-Hodgkin lymphoma.

Recently there has been a growing interest in evaluating body composition as a marker for prognosis in cancer patients. The association of body composition parameters and outcomes has not been deeply investigated in patients with autologous hematopoietic stem cell transplantation (HSCT) recipients with non-Hodgkin lymphoma (NHL). We conducted a retrospective cohort study of 264 NHL patients who received autologous HSCT. PreHSCT abdominal CT scans at the levels of L3 were assessed for body composition measures. We evaluated sarcopenia, myosteatosis, high visceral adipose tissue (VAT) and high visceral adipose tissue density (VATD). Using multivariable Cox proportional regression, we analyzed the association of clinical and transplant-related characteristics with overall survival (OS), relapse-free survival (RFS), and non-relapse mortality (NRM). In a multivariate regression model, patients with higher VATD had worse OS (HR 1.78; 95% confidence intervals CI 1.08-2.95, p = 0.02) and worse NRM (HR 2.31 95% CI 1.08-4.95, p = 0.02) than with lower VATD. Patients with lower levels of VAT also had worse RFS (HR 1.49 95% CI 1.03-2.15, p = 0.03). Sarcopenia and myosteatosis were not associated with outcomes. High pre-transplant VATD was associated with lower OS and higher NRM, and low pre-transplant VAT was associated with worse RFS in patients with NHL undergoing autologous HSCT.

Climbing strategies of Taiwan climbers.

BACKGROUND: The climbing strategies of lianas and herbaceous vines influence climber competition abilities and survival. The aim of this study was to investigate the climbing strategies of each plant species and observe their organs of origin. RESULTS: The results showed that all Taiwan climbers were approximately 555 species, accounting for 11% of the native flora. Among the 555 climbers, the twining stem type was the most common, with a total of 255 species (46%), the remaining climbing methods accounted for 300 species. Approximately twenty one climbing methods, including nine combination types, were exhibited, of which the most common type was the twining stem, followed by simple scrambling and twining tendrils. Most species of Fabaceae and Apocynaceae were twining stems in dextrorse, excluding Wisteriopsis reticulata and Alyxia taiwanensis, which were in sinistrorse. The prehensile branch of Fissistigma genus, Ventilago genus, and Dalbergia benthamii, originated from second-order or modified stems. In the simple scrambling type, some climbers were covered spines and prickles to attach the host, and the others were clinging to the supports or creeping on the ground without speculation. The hooks or grapnels of the genus Uncaria are derived from the branches, and a pair of curved hooks or a spine of Artabotrys hexapetalus are originated from the inflorescence to tightly attach to a host. The Piper genus use adhesive roots to climb their hosts. Among the genus Trichosanthes, only Trichosanthes homophylla exhibits a combination of twining modified shoots and adhesive roots. Gentianales includes four families with seven climbing mechanisms, while Fabales includes only Fabaceae, which presents six climbing methods. CONCLUSIONS: The twining tendrils had nine organs of origin in Taiwan climber, that these opinions of originated organs might be available to the studies of convergent evolution. The data presented herein provide crucial basic information of the climber habits types and origin structures, which are available for terms standardization to improve field investigation. The terminologies would aid in the establishment of climber habits as commonly taxon-specific and the combination of two climber habits could be a characteristic of taxonomic value.

Significance of Adipose Tissue Quantity and Distribution on Obesity Paradox in Heart Failure.

Obesity is a predictor of the development of systolic and diastolic heart failure (HF), but once established, patients with HF and obesity have better outcomes than their leaner counterparts, a phenomenon termed the "obesity paradox." We sought to investigate the impact of adipose tissue quantity and distribution, measured by way of computed tomography, on outcomes in patients with HF. Patients admitted for acute decompensated HF between January 2017 to December 2018 were retrospectively analyzed. Body composition measurements were made on computed tomography of the abdomen/pelvis. Visceral, subcutaneous, and intermuscular adipose tissues were measured at the mid-third lumbar vertebra, along with skeletal muscle and waist circumference. Paracardial (pericardial and epicardial) adipose tissue was measured at the mid-eight thoracic vertebra. Visceral adipose tissue index (VATI) and subcutaneous adipose tissue index (SATI), along with skeletal muscle index, were indexed for patient height. A total of 200 patients were included, 44.5% female. Body mass index and waist circumference did not significantly predict outcomes. Patients with high SATI (highest sex-stratified tertile) had significantly better survival (hazard ratio 0.58, 95% confidence interval 0.39 to 0.87, p = 0.009), whereas high VATI was nonsignificant. Patients were further divided into 4 groups based on both VATI and SATI. One- and 4-year mortality risks were lowest in those with low VATI high SATI compared with the other groups; this persisted after multivariable adjustment for covariates, including albumin and skeletal muscle index. In conclusion, the "obesity paradox" appears to be largely driven by subcutaneous adipose tissue, independent of nutrition or skeletal muscle.

Sarcopenia evaluation on cardiac magnetic resonance imaging in older adults for outcomes prediction following surgical aortic valve replacement.

BACKGROUND: Sarcopenia refers to a reduction in skeletal muscle mass and strength. Despite the known association between single-slice muscle measurements on lumbar computed tomography and poor outcomes in various clinical settings, studies using thoracic muscle measurements on cardiac magnetic resonance imaging (CMR) have been limited. METHODS: Patients undergoing surgical aortic valve replacement (SAVR) between 2010 and 2020 were included if they were ≥ 50 years of age with preoperative CMR. Manual unilateral pectoralis major and minor skeletal muscle area measurements were made at the carina and normalized for body size by height to obtain skeletal muscle index (SMI). Sarcopenia was defined as the lowest sex-stratified SMI tertile and higher-risk as the highest fiftieth percentile Society of Thoracic Surgeons' (STS) mortality score. RESULTS: A total of 133 patients were included, 35 (26.3%) females. The average age was 64 ± 9 years, with most Caucasian (93.2%). Compared to non-sarcopenic patients, sarcopenic patients were older with lower body mass index. During a median follow-up of 27.3 (7.6-60.4) months, 10 (22.2%) deaths occurred in the sarcopenic group and 8 (9.1%) in the non-sarcopenic group (p = 0.039 by log-rank test). On subgroup analysis (66 patients), higher-risk sarcopenic patients had 10 (37.0%) deaths compared to 8 (20.5%) in higher-risk non-sarcopenic patients (p = 0.011 by log-rank test). CONCLUSIONS: Simple unilateral pectoralis muscle measurements on preoperative CMR can be used as an adjunct to traditional risk scores for predicting mortality post-SAVR.

Keeping Patient Data Secure in the Age of Radiology Artificial Intelligence: Cybersecurity Considerations and Future Directions.

Artificial intelligence (AI)-based solutions are increasingly being incorporated into radiology workflows. Implementation of AI comes along with cybersecurity risks and challenges that practices should be aware of and mitigate for a successful and secure deployment. In this article, these cybersecurity issues are examined through the lens of the "CIA" triad framework-confidentiality, integrity, and availability. We discuss the implications of implementation configurations and development approaches on data security and confidentiality and the potential impact that the insertion of AI can have on the truthfulness of data, access to data, and the cybersecurity attack surface. Finally, we provide a checklist to address important security considerations before deployment of an AI application, and discuss future advances in AI addressing some of these security concerns.

Harnessing microbial phylum-specific molecular markers for assessment of environmental estrogen degradation.

Steroidal estrogens are ubiquitous contaminants that have garnered attention worldwide due to their endocrine-disrupting and carcinogenic activities at sub-nanomolar concentrations. Microbial degradation is one of the main mechanisms through which estrogens can be removed from the environment. Numerous bacteria have been isolated and identified as estrogen degraders; however, little is known about their contribution to environmental estrogen removal. Here, our global metagenomic analysis indicated that estrogen degradation genes are widely distributed among bacteria, especially among aquatic actinobacterial and proteobacterial species. Thus, by using the Rhodococcus sp. strain B50 as the model organism, we identified three actinobacteria-specific estrogen degradation genes, namely aedGHJ, by performing gene disruption experiments and metabolite profile analysis. Among these genes, the product of aedJ was discovered to mediate the conjugation of coenzyme A with a unique actinobacterial C17 estrogenic metabolite, 5-oxo-4-norestrogenic acid. However, proteobacteria were found to exclusively adopt an α-oxoacid ferredoxin oxidoreductase (i.e., the product of edcC) to degrade a proteobacterial C18 estrogenic metabolite, namely 3-oxo-4,5-seco-estrogenic acid. We employed actinobacterial aedJ and proteobacterial edcC as specific biomarkers for quantitative polymerase chain reaction (qPCR) to elucidate the potential of microbes for estrogen biodegradation in contaminated ecosystems. The results indicated that aedJ was more abundant than edcC in most environmental samples. Our results greatly expand the understanding of environmental estrogen degradation. Moreover, our study suggests that qPCR-based functional assays are a simple, cost-effective, and rapid approach for holistically evaluating estrogen biodegradation in the environment.

Identification of acute aortic syndromes based on cross-sectional variability of Hounsfield units.

BACKGROUND: A characteristic feature of communicating aortic dissections (CD) is the dissection flap between the true and false lumen. However, in intramural hematomas (IMH) a flap is not visible. We aimed to determine if cross-sectional HU variability allow reliable identification of aortic dissections including IMH. METHODS: We included 362 patients presenting with acute chest pain (CP) or respiratory distress (RD) and underwent contrast-enhanced CTA with or without ECG-gating. In the derivation group we included 72 CP patients with and 74 without AAS. In the validation group we included 108 CP or RD patients with and 108 without AAS. The adventitial border of the aorta was visually identified and measurements were performed at 6 locations along the ascending and descending aorta. At each cross-section 5 circular ROI measurements of HU were made and the maximum HU difference calculated. RESULTS: In the derivation and validation group the maximum difference in HUs at any one location was significantly higher for AAS subjects than controls (validation group: median = 128.5 vs. 34.0, p-value Wilcoxon two-sample test <0.001). In the validation group, the estimated AUC was 0.939 with 95% CIs of [0.906, 0.972], indicating that the maximum difference in HUs is a strong predictor of AAS (p < 0.001). CONCLUSION: Our data provide evidence that cross-sectional variability of Hounsfield Unit reliably identifies aortic dissection including IMH in dedicated ECG-gated aorta scans but also non-gated chest CTs with limited aortic contrast enhancement. These results suggest that this approach could be feasible for an automated algorithm for identification of AAS.

Circulating androgen regulation by androgen-catabolizing gut bacteria in male mouse gut.

Abnormally high circulating androgen levels have been considered a causative factor for benign prostatic hypertrophy and prostate cancer in men. Recent animal studies on gut microbiome suggested that gut bacteria are involved in sex steroid metabolism; however, the underlying mechanisms and bacterial taxa remain elusive. Denitrifying betaproteobacteria Thauera spp. are metabolically versatile and often distributed in the animal gut. Thauera sp. strain GDN1 is an unusual betaproteobacterium capable of catabolizing androgen under both aerobic and anaerobic conditions. We administered C57BL/6 mice (aged 7 weeks) with strain GDN1 through oral gavage. The strain GDN1 administration caused a minor increase in the relative abundance of Thauera (≤0.1%); however, it has profound effects on the host physiology and gut bacterial community. The results of our ELISA assay and metabolite profile analysis indicated an approximately 50% reduction in serum androgen levels in the strain GDN1-administered male mice. Moreover, androgenic ring-cleaved metabolites were detected in the fecal extracts of the strain GDN1-administered mice. Furthermore, our RT - qPCR results revealed the expression of the androgen catabolism genes in the gut of the strain GDN1-administered mice. We found that the administered strain GDN1 regulated mouse serum androgen levels, possibly because it blocked androgen recycling through enterohepatic circulation. This study discovered that sex steroids serve as a carbon source of gut bacteria; moreover, host circulating androgen levels may be regulated by androgen-catabolizing gut bacteria. Our data thus indicate the possible applicability of androgen-catabolic gut bacteria as potent probiotics in alternative therapy of hyperandrogenism.

Sarcopenia detected with bioelectrical impedance versus CT scan and chemotherapy tolerance in patients with early breast cancer.

BACKGROUND: Early breast cancer (EBC) is a potentially curable disease. Some patients with EBC require chemotherapy, exposing patients to undesirable side effects. Loss of muscle mass, or sarcopenia, has been associated with worse outcomes in patients with EBC and worse treatment-related toxicity in patients with advanced breast cancer. CT scans can identify sarcopenia; however, most patients with EBC do not require routine CT scans. Bioelectrical impedance spectrometry (BIS) is another method to detect sarcopenia and can be performed quickly in the office without radiation exposure. We sought to investigate whether sarcopenia measurements by CT scan versus BIS correlated with each other and whether sarcopenia identified by each method is associated with chemotherapy toxicity and adherence in patients with EBC. METHODS: This is a retrospective study; eligible patients received chemotherapy treatment for EBC and had undergone BIS. A subset of patients had also had a CT abdomen with a Lumbar L3 level. Measures of sarcopenia were obtained from the BIS and CT data. In addition, patient characteristics, treatment, and toxicity-related outcomes were obtained from medical records. Multivariate logistic regression models were used to associate sarcopenia status with toxicity endpoints, adjusted for other patient characteristics. RESULTS: There was a moderate correlation between sarcopenia detected by CT scan and BIS (r = 0.64 p < 0.0001). Patients with sarcopenia detected by BIS had more chemotherapy toxicity (OR = 2.56; CI 1.72-3.84), dose reductions or dose delays (OR = 1.58; CI 1.06-2.38), and hospitalizations (OR = 2.38; CI 1.33-4.16) due to side effects than patients without sarcopenia. CONCLUSION: The presence of sarcopenia in patients with EBC is associated with worse chemotherapy tolerance. BIS represents a high-value alternative to CT scans for sarcopenia assessment.

Stem cambial variants of Taiwan lianas.

BACKGROUND: Cambium in lianas, responsible for secondary growth, develop diverse and diagnostic traits during the climbing phase. Studies on the cross-section of Taiwanese liana cambial variants are scarce. We collected multiple stem cross-sections from 287 liana species belonging to 52 families. Each sample was examined on five occasions, and the observations were documented. RESULTS: The results showed that approximately 22 cambial variants types were displayed in Taiwan lianas. Among these, axial vascular elements in radial segments were the most common, followed by the variants with the irregular conformation and intraxylary phloem. Based on our assessment, we provide the following identification features of a few families: Apocynaceae had intraxylary phloem; Convolvulaceae had intraxylary phloem combined with successive cambia; Lardizabalaceae, Menispermaceae, and Ranunculaceae possessed axial vascular elements in segments; Piperaceae had external primary vascular bundle cylinder combined with axial vascular elements in segments; Vitaceae had axial vascular elements in segments combined with irregular conformation. Axial vascular elements in segments and intraxylary phloem appeared in six or five combination types, showing that these two types combined with many variants are helpful for the identification of lianas. Two species, Momordica charantia var. abbreviata, and Momordica cochinchinensis had a cambium element in the outer cylinder of cortical bicollateral vascular bundles and formed directional layers of successive cambia. CONCLUSIONS: Our study documented regular secondary growth with a single cambium in 36 species and cambial variants present in 16 species of Taiwanese lianas. Furthermore, we provide crucial baseline data on liana cambial variations, thereby improving our understanding of their morphology and identification.

Current Approach to the Diagnosis of Sarcopenia in Heart Failure: A Narrative Review on the Role of Clinical and Imaging Assessments.

Sarcopenia has been established as a predictor of poor outcomes in various clinical settings. It is particularly prevalent in heart failure, a clinical syndrome that poses significant challenges to health care worldwide. Despite this, sarcopenia remains overlooked and undertreated in cardiology practice. Understanding the currently proposed diagnostic process is paramount for the early detection and treatment of sarcopenia to mitigate downstream adverse health outcomes.

Natural Language Processing for Imaging Protocol Assignment: Machine Learning for Multiclass Classification of Abdominal CT Protocols Using Indication Text Data.

A correct protocol assignment is critical to high-quality imaging examinations, and its automation can be amenable to natural language processing (NLP). Assigning protocols for abdominal imaging CT scans is particularly challenging given the multiple organ specific indications and parameters. We compared conventional machine learning, deep learning, and automated machine learning builder workflows for this multiclass text classification task. A total of 94,501 CT studies performed over 4 years and their assigned protocols were obtained. Text data associated with each study including the ordering provider generated free text study indication and ICD codes were used for NLP analysis and protocol class prediction. The data was classified into one of 11 abdominal CT protocol classes before and after augmentations used to account for imbalances in the class sample sizes. Four machine learning (ML) algorithms, one deep learning algorithm, and an automated machine learning (AutoML) builder were used for the multilabel classification task: Random Forest (RF), Tree Ensemble (TE), Gradient Boosted Tree (GBT), multi-layer perceptron (MLP), Universal Language Model Fine-tuning (ULMFiT), and Google's AutoML builder (Alphabet, Inc., Mountain View, CA), respectively. On the unbalanced dataset, the manually coded algorithms all performed similarly with F1 scores of 0.811 for RF, 0.813 for TE, 0.813 for GBT, 0.828 for MLP, and 0.847 for ULMFiT. The AutoML builder performed better with a F1 score of 0.854. On the balanced dataset, the tree ensemble machine learning algorithm performed the best with an F1 score of 0.803 and a Cohen's kappa of 0.612. AutoML methods took a longer time for completion of NLP model training and evaluation, 4 h and 45 min compared to an average of 51 min for manual methods. Machine learning and natural language processing can be used for the complex multiclass classification task of abdominal imaging CT scan protocol assignment.