-
Journal of the American College of... Jun 2024Thyroid nodule evaluation using ultrasound is dependent on radiologist experience, but deep learning (DL) models can improve intra-reader agreements. DL model...
Thyroid nodule evaluation using ultrasound is dependent on radiologist experience, but deep learning (DL) models can improve intra-reader agreements. DL model development for medical imaging with small datasets can be challenging. Transfer learning is a technique used in the development of DL models to improve model performance in data-limited scenarios. Here, we investigate the impact of transfer learning with domain-specific RadImageNet dataset and non-medical ImageNet on the robustness of classifying thyroid nodules into benign and malignant. We retrospectively collected 822 ultrasound images of thyroid nodules of patients who underwent fine needle aspiration in our institute. We split our data and used 101 cases in a test set and 721 cases for cross-validation. A Resnet-18 model was trained to classify thyroid nodules into benign and malignant. Then, we trained the same model architecture with transferred weights from ImageNet and RadImageNet. The model without transfer learning for thyroid nodule classification achieved an AUROC of 0.69. The AUROC of our model after transfer learning with ImageNet pre-trained weights was 0.79. Our model achieved an AUROC of 0.83 from transfer learning of the RadImageNet pre-trained weights. The AUROC from the classification model without transfer learning significantly improved after transfer learning with ImageNet (p-value = 0.03) and RadImageNet transfer learning (p-value <0.01). There was a statistically significant distinction in performance between the model utilizing RadImageNet transfer learning and that employing ImageNet transfer learning (p-value <0.01). We demonstrate the potential of RadImageNet as a domain-specific source for transfer learning in thyroid nodule classification.
PubMed: 38942163
DOI: 10.1016/j.jacr.2024.06.011 -
Maturitas Jun 2024Cross-sectional evidence has shown that frailty is highly prevalent in patients with chronic kidney disease (CKD). However, there is limited evidence of the longitudinal...
OBJECTIVES
Cross-sectional evidence has shown that frailty is highly prevalent in patients with chronic kidney disease (CKD). However, there is limited evidence of the longitudinal associations between frailty, genetic predisposition to CKD, and the risk of CKD in the general population. Therefore, this study aimed to examine such associations among participants in the UK Biobank.
STUDY DESIGN
This is a prospective cohort study included 370,965 middle-aged and older adults from the UK Biobank. Physical frailty was assessed using a modified version of the Fried phenotype classification. A weighted genetic risk score was built using 263 variants associated with estimated glomerular filtration rate.
MAIN OUTCOME MEASURES
Incident CKD was identified from hospital inpatient records.
RESULTS
Over a median follow-up of 12.3 years, we documented a total of 11,121 incident CKD cases. Time-dependent Cox proportional hazards regression models indicated that individuals with frailty (hazard ratio [HR]: 1.94, 95 % confidence interval [CI]: 1.81-2.08) and pre-frailty (HR: 1.27, 95 % CI: 1.22-1.33) had an increased risk of developing CKD, compared with non-frail individuals. No significant interaction between frailty and genetic risk score was observed (P for interaction = 0.41). The highest risk was observed among the individuals with high genetic risk and frailty (HR: 2.31, 95 % CI: 2.00-2.68).
CONCLUSION
Our results demonstrated that frailty and pre-frailty were associated with increased risk of incident CKD in middle-age and older adults, regardless of genetic risk of CKD. Our study underscores the importance of frailty screening and intervention as a potential strategy to prevent CKD. Future clinical trials are needed to validate our findings.
PubMed: 38941958
DOI: 10.1016/j.maturitas.2024.108059 -
Artificial Intelligence in Medicine Jun 2024Pancreatic cancer does not show specific symptoms, which makes the diagnosis of early stages difficult with established image-based screening methods and therefore has...
Pancreatic cancer does not show specific symptoms, which makes the diagnosis of early stages difficult with established image-based screening methods and therefore has the worst prognosis among all cancers. Although endoscopic ultrasonography (EUS) has a key role in diagnostic algorithms for pancreatic diseases, B-mode imaging of the pancreas can be affected by confounders such as chronic pancreatitis, which can make both pancreatic lesion segmentation and classification laborious and highly specialized. To address these challenges, this work proposes a semi-supervised multi-task network (SSM-Net) to leverage unlabeled and labeled EUS images for joint pancreatic lesion classification and segmentation. Specifically, we first devise a saliency-aware representation learning module (SRLM) on a large number of unlabeled images to train a feature extraction encoder network for labeled images by computing a contrastive loss with a semantic saliency map, which is obtained by our spectral residual module (SRM). Moreover, for labeled EUS images, we devise channel attention blocks (CABs) to refine the features extracted from the pre-trained encoder on unlabeled images for segmenting lesions, and then devise a merged global attention module (MGAM) and a feature similarity loss (FSL) for obtaining a lesion classification result. We collect a large-scale EUS-based pancreas image dataset (LS-EUSPI) consisting of 9,555 pathologically proven labeled EUS images (499 patients from four categories) and 15,500 unlabeled EUS images. Experimental results on the LS-EUSPI dataset and a public thyroid gland lesion dataset show that our SSM-Net clearly outperforms state-of-the-art methods.
PubMed: 38941908
DOI: 10.1016/j.artmed.2024.102919 -
Journal of Biomechanics Jun 2024For some individuals with severe socket-related problems, prosthesis osseointegration directly connects a prosthesis to the residual limb creating a bone-anchored limb...
For some individuals with severe socket-related problems, prosthesis osseointegration directly connects a prosthesis to the residual limb creating a bone-anchored limb (BAL). We compared dynamic gait stability and between-limb stability symmetry, as measured by the Margin of Stability (MoS) and the Normalized Symmetry Index (NSI), for people with unilateral transfemoral amputation before and one-year after BAL implantation. The MoS provides a mechanical construct to assess dynamic gait stability and infer center of mass and limb control by relating the center of mass and velocity to the base of support. Before and one-year after BAL implantation, 19 participants walked overground at self-selected speeds. We quantified dynamic gait stability anteriorly and laterally at foot strike and at the minimum lateral MoS value. After implantation, we observed decreased lateral MoS at foot strike for the amputated (MoS mean(SD) %height; pre: 6.6(2.3), post: 5.9(1.3), d = 0.45) and intact limb (pre: 6.2(1.2), post: 5.8(1.0), d = 0.38) and increased between-limb MoS symmetry at foot strike (NSI mean(SD) %; anterior-pre: 10.3(7.3), post: 8.4(3.6), d = 0.23; lateral-pre: 18.8(12.4), post: 12.4(4.9), d = 0.47) and at minimum lateral stability (pre: 28.1(18.1), post: 19.2(6.8), d = 0.50). Center of mass control using a BAL resulted in dynamic gait stability more similar between limbs and may have reduced the adoption of functional asymmetries. We suggest that improved between-limb MoS symmetry after BAL implantation is likely due to subtle changes in individual limb MoS values at self-selected walking speeds resulting in an overall positive impact on fall risk through improved center of mass and prosthetic limb control.
PubMed: 38941842
DOI: 10.1016/j.jbiomech.2024.112208 -
General Hospital Psychiatry Jun 2024Several medications are associated with delirium; however, studies with adequate statistical power are limited, and it is difficult to determine the effects of the...
OBJECTIVE
Several medications are associated with delirium; however, studies with adequate statistical power are limited, and it is difficult to determine the effects of the various concomitant medications used in clinical practice. Therefore, in this study, we aimed to comprehensively evaluate the safety signals of delirium-associated drugs using a spontaneous adverse event reporting system.
METHOD
The JAPIC AERS (Food and Drug Administration Adverse Event Reporting System pre-processed by the Japan Pharmaceutical Information Center) was used for the analysis in this pharmacovigilance study. The reporting odds ratio (ROR) for delirium was adjusted for using multivariate logistic regression analysis with sex, age, indication, and melatonin receptor agonist use, and 22 drug categories were targeted as covariates.
RESULTS
After excluding patients with missing information, 7,527,568 patients were included in the study. Delirium signals were detected even after adjusting for covariates in 17 drug categories, including benzodiazepines (adjusted ROR, 1.76; 95% confidence interval [CI], 1.64-1.89), opioids (adjusted ROR, 4.42; 95% CI, 4.21-4.64), and tricyclic antidepressants (adjusted ROR, 2.44; 95% CI, 2.20-2.71).
CONCLUSIONS
These findings suggest that many drug classes, such as benzodiazepines, are independent risk factors for delirium and strengthen the evidence of an association between delirium and medications.
PubMed: 38941744
DOI: 10.1016/j.genhosppsych.2024.06.012 -
International Journal of Pediatric... Jun 2024There is limited but consistent evidence that suggests prenatal factors, including maternal stress, may contribute to susceptibility for otitis media. We aimed to...
OBJECTIVES
There is limited but consistent evidence that suggests prenatal factors, including maternal stress, may contribute to susceptibility for otitis media. We aimed to determine the effect of multiple life stress events during pregnancy on risk of acute and recurrent otitis media in offspring at three and five years of age.
METHODS
Exposure data on stressful life events were collected from pregnant women in a longitudinal prospective pregnancy cohort study, at 18 and 34 weeks' gestation. We used longitudinal regression models stratified by offspring sex to examine associations between the number, type and timing of maternal prenatal stress events and the likelihood of any OM in addition to recurrent OM infection at age three and five years, adjusting for pre-specified prenatal sociodemographic and environmental confounders.
RESULTS
Each additional stressful life event in pregnancy was associated with increased risk of any OM at both ages (3 years: OR = 1.07, 95%CI = 1.02, 1.12; 5 years: OR = 1.07, 95%CI = 1.02, 1.12), with larger effect sizes for recurrent otitis media (3 years: OR = 1.11, 95%CI = 1.05, 1.17; 5 years: OR = 1.09, 95%CI = 1.04, 1.14). Risk of offspring otitis media did not differ with timing of stress nor by offspring sex. Specific types of stress (pregnancy and relationship problems, issues with other children) were each associated with increased risk of recurrent OM at age three and five years.
CONCLUSIONS
We observed a dose-response relationship between maternal stressful life events in pregnancy and the risk for offspring otitis media in the preschool years, most marked for recurrent otitis media.
PubMed: 38941719
DOI: 10.1016/j.ijporl.2024.112022 -
The Journal of Surgical Research Jun 2024Glucagon-like peptide-1 receptor agonist (GLP-1A) medications are gaining widespread popularity for the treatment of obesity. The optimal use of these drugs in pediatric...
INTRODUCTION
Glucagon-like peptide-1 receptor agonist (GLP-1A) medications are gaining widespread popularity for the treatment of obesity. The optimal use of these drugs in pediatric bariatric populations, and especially in those considering metabolic and bariatric surgery (MBS), is yet to be established. We sought to characterize current practice patterns of GLP-1A use at major pediatric bariatric centers across the United States.
MATERIALS AND METHODS
We administered an online survey to a purposive sample of 46 surgeons who perform MBS on children and adolescents. Survey questions explored practices prescribing GLP-1As in patients considering MBS, holding them prior to elective operations, and restarting them postoperatively following MBS. Responses were summarized with descriptive statistics and inductive content analysis.
RESULTS
There were 22 responses (48% response rate) representing 19 institutions. Most (86%) respondents do sometimes prescribe GLP-1As for patients considering MBS, but the specific indications vary. Practices for holding GLP-1As preoperatively also vary, from not at all to holding for 2 wk. Over half (55%) of respondents sometimes restart GLP-1As after MBS. Free-response themes included still-evolving preoperative utilization patterns, difficulty with access and insurance coverage, and a lack of data informing GLP-1A use in the pre and postoperative periods.
CONCLUSIONS
Given the increasing use of these medications for weight loss purposes, this substantial variation in practice highlights a need for further research to examine the safest and most effective use of GLP-1As in the pre and postoperative periods and for practice guidelines to standardize care pathways in pediatric bariatric contexts.
PubMed: 38941713
DOI: 10.1016/j.jss.2024.05.045 -
Public Health Jun 2024COVID-19 revealed major shortfalls in healthcare workers (HCWs) trained in acute and critical care worldwide, especially in low-resource settings. We aimed to assess...
OBJECTIVES
COVID-19 revealed major shortfalls in healthcare workers (HCWs) trained in acute and critical care worldwide, especially in low-resource settings. We aimed to assess mass online courses' efficacy in preparing HCWs to manage COVID-19 patients and to determine whether rapidly deployed e-learning can enhance their knowledge and confidence during a pandemic.
STUDY DESIGN
Retrospective cohort study.
METHODS
This international retrospective cohort study, led by a large Academic Medical Centre (AMC), was conducted via YouTube and the AMC's online learning platform. From 2020 to 2021, multidisciplinary experts developed and deployed six online training courses based on the latest evidence-based management guidelines. Participants were selected through a voluntary sample following an electronic campaign. Training outcomes were assessed using pre-and post-test questionnaires, evaluation forms, and post-training assessment surveys. Kirkpatrick's Model guided training evaluation to measure self-reported knowledge, clinical skills, and confidence improvement. We also captured the number and type of COVID-19 patients managed by HCWs after the trainings.
RESULTS
Every 22.8 reach/impression and every 1.2 engagements led to a course registration. The 10,425 registrants (56.8% female, 43.1% male) represented 584 medical facilities across 154 cities. The largest segments of participants were students/interns (20.6%) and medical officers (13.4%). Of the 2169 registered participants in courses with tests, 66.9% completed post-tests. Test scores from all courses increased from the initial baseline to subsequent improvement post-course. Participants completing post-training assessment surveys reported that the online courses improved their knowledge and clinical skills (83.5%) and confidence (89.4%). Respondents managed over 19,720 COVID-19 patients after attending the courses, with 47.7% patients being moderately/severely ill.
CONCLUSIONS
Participants' confidence in handling COVID-19 patients is increased by rapidly deploying mass training to a substantial target population through digital tools. The findings present a virtual education and assessment model that can be leveraged for future global public health issues, and estimates for future electronic campaigns to target.
PubMed: 38941682
DOI: 10.1016/j.puhe.2024.05.006 -
Journal of Neurosurgery Jun 2024Recent work on ischemic cerebellar stroke has suggested that the resection of infarcted tissue may lead to improved functional outcomes compared with decompressive...
OBJECTIVE
Recent work on ischemic cerebellar stroke has suggested that the resection of infarcted tissue may lead to improved functional outcomes compared with decompressive surgery alone. Nonetheless, no studies have assessed the extent to which necrotic tissue should be resected or if there are any volumetric thresholds capable of predicting functional outcomes in this patient population. In this study, the authors aimed to determine potential thresholds for volume reduction in ischemic cerebellar stroke in an effort to optimize the management of ischemic cerebellar stroke and, in so doing, improve functional outcomes.
METHODS
This study is a multicentric retrospective study of patients who underwent surgery for the management of ischemic cerebellar stroke. Volumetric analyses of infarcted tissue present on CT scans were performed before and after surgical intervention(s). The final infarct volume (FIV) was computed as a percentage of the initial infarct volume (postoperative infarct volume/preoperative infarct volume × 100). The primary endpoint was functional outcome at 3 months, as determined by the modified Rankin Scale (mRS) score; mRS scores 0-2 were considered as favorable and mRS scores 3-6 as unfavorable. Receiver operating characteristic curves were used to explore the relationship between postoperative infarct volumes and FIV versus mRS score, and Youden's index was used to estimate potential volumetric thresholds.
RESULTS
A total of 91 patients were included in the study. The mean pre- and postoperative infarct volumes were 45.25 (SD 18.32) cm3 and 29.56 (SD 26.61) cm3, respectively. Patients undergoing necrosectomy, regardless of whether via craniotomy or craniectomy, were more likely to have a favorable outcome at discharge (OR 16.62, 95% CI 2.12-130.33; p = 0.008) and at 3 months (OR 24.12, 95% CI 3.03-192.18; p = 0.003) postoperatively. Postoperative infarct volumes ≤ 17 cm3 yielded a sensitivity of 77% and a specificity of 68% with regard to the prediction of favorable outcome at 3 months. The resection ≥ 50% of infarcted tissue was also predictive of favorable outcomes at 3 months (OR 7.7, 95% CI 2.7-21.8; p < 0.001).
CONCLUSIONS
The reduction of necrotic tissue volumes by at least 50% and/or the reduction of the infarct volume by ≤ 17 cm3 appear to be associated with favorable outcomes in patients with surgically managed ischemic cerebellar strokes.
PubMed: 38941630
DOI: 10.3171/2024.3.JNS232883 -
Research Quarterly For Exercise and... Jun 2024The present study investigated the effects of a 10-week preparatory training period on biomarkers and jumping performance and associations of changes in biomarkers,...
Anabolic/Catabolic Hormone Imbalance but Still Jumping Further? Negative Association of Free Testosterone With Jumping Performance in Elite Handball Players Following a Preparatory Period.
The present study investigated the effects of a 10-week preparatory training period on biomarkers and jumping performance and associations of changes in biomarkers, load, and jumping performance from the beginning (PRE) to the end of the preparatory period (POST) in elite handball players. Seventeen elite handball players competing in the first Slovenian men's League were recruited. Training, competition and academic loads were reported weekly, while biomarkers and jumping performance were assessed at PRE and POST. At POST, decreased levels of free testosterone (large effect size [ES] = -1.69, < .001) and free testosterone to cortisol ratio [FTCR] (large ES = -.95, = .004) were observed; whereas, better performance on the single leg lateral hop test [SLLH] (large ES = .85, = .007) and single leg triple hop test [SLTH] (large ES = 1.05, = .002) were observed compared to PRE. Furthermore, changes in FTCR correlated with changes in cortisol (high = -.751, = .001), SLLH (moderate = -.603, = .022), and SLTH (moderate = -.643, = .013), while changes in free testosterone correlated with SLTH (moderate = -.645, = .013). High intensity trainings with a saturated competition schedule can result in disturbed anabolic/catabolic hormone ratio observed through FTCR decrease, which could indicate either an optimal state or early exhaustiveness. It seems that SLLH and SLTH are more sensitive to changes in biomarkers than a single leg hop test. Sport professionals may use the results for individualized monitoring of an athlete's health and performance, specifically, as an aid for adjusting training loads accordingly to prevent performance declines and potential injury/illness events.
PubMed: 38941621
DOI: 10.1080/02701367.2024.2353715