Tag: nevin manimala

Am Nat. 2024 Aug;204(2):105-120. doi: 10.1086/730807. Epub 2024 Jun 20.

ABSTRACT

AbstractInteractions between and within abiotic and biotic processes generate nonadditive density-dependent effects on species performance that can vary in strength or direction across environments. If ignored, nonadditivities can lead to inaccurate predictions of species responses to environmental and compositional changes. While there are increasing empirical efforts to test the constancy of pairwise biotic interactions along environmental and compositional gradients, few assess both simultaneously. Using a nationwide forest inventory that spans broad ambient temperature and moisture gradients throughout New Zealand, we address this gap by analyzing the diameter growth of six focal tree species as a function of neighbor densities and climate, as well as neighbor × climate and neighbor × neighbor statistical interactions. The most complex model featuring all interaction terms had the highest predictive accuracy. Compared with climate variables, biotic interactions typically had stronger effects on diameter growth, especially when subjected to nonadditivities from local climatic conditions and the density of intermediary species. Furthermore, statistically strong (or weak) nonadditivities could be biologically irrelevant (or significant) depending on whether a species pair typically interacted under average or more extreme conditions. Our study highlights the importance of considering both the statistical potential and the biological relevance of nonadditive biotic interactions when assessing species performance under global change.

PMID:39008837 | DOI:10.1086/730807

J Med Internet Res. 2024 Jul 15;26:e57762. doi: 10.2196/57762.

ABSTRACT

BACKGROUND: Early detection of cancer and provision of appropriate treatment can increase the cancer cure rate and reduce cancer-related deaths. Early detection requires improving the cancer screening quality of each medical institution and enhancing the capabilities of health professionals through tailored education in each field. However, during the COVID-19 pandemic, regional disparities in educational infrastructure emerged, and educational accessibility was restricted. The demand for remote cancer education services to address these issues has increased, and in this study, we considered medical metaverses as a potential means of meeting these needs. In 2022, we used Metaverse Educational Center, developed for the virtual training of health professionals, to train radiologic technologists remotely in mammography positioning.

OBJECTIVE: This study aims to investigate the user experience of the Metaverse Educational Center subplatform and the factors associated with the intention for continuous use by focusing on cases of using the subplatform in a remote mammography positioning training project.

METHODS: We conducted a multicenter, cross-sectional survey between July and December 2022. We performed a descriptive analysis to examine the Metaverse Educational Center user experience and a logistic regression analysis to clarify factors closely related to the intention to use the subplatform continuously. In addition, a supplementary open-ended question was used to obtain feedback from users to improve Metaverse Educational Center.

RESULTS: Responses from 192 Korean participants (male participants: n=16, 8.3%; female participants: n=176, 91.7%) were analyzed. Most participants were satisfied with Metaverse Educational Center (178/192, 92.7%) and wanted to continue using the subplatform in the future (157/192, 81.8%). Less than half of the participants (85/192, 44.3%) had no difficulty in wearing the device. Logistic regression analysis results showed that intention for continuous use was associated with satisfaction (adjusted odds ratio 3.542, 95% CI 1.037-12.097; P=.04), immersion (adjusted odds ratio 2.803, 95% CI 1.201-6.539; P=.02), and no difficulty in wearing the device (adjusted odds ratio 2.020, 95% CI 1.004-4.062; P=.049). However, intention for continuous use was not associated with interest (adjusted odds ratio 0.736, 95% CI 0.303-1.789; P=.50) or perceived ease of use (adjusted odds ratio 1.284, 95% CI 0.614-2.685; P=.51). According to the qualitative feedback, Metaverse Educational Center was useful in cancer education, but the experience of wearing the device and the types and qualities of the content still need to be improved.

CONCLUSIONS: Our results demonstrate the positive user experience of Metaverse Educational Center by focusing on cases of using the subplatform in a remote mammography positioning training project. Our results also suggest that improving users’ satisfaction and immersion and ensuring the lack of difficulty in wearing the device may enhance their intention for continuous use of the subplatform.

PMID:39008834 | DOI:10.2196/57762

J Med Internet Res. 2024 Jul 15;26:e51814. doi: 10.2196/51814.

ABSTRACT

BACKGROUND: Telepsychiatry (TP), a live video meeting, has been implemented in many contexts and settings. It has a distinct advantage in the psychiatric emergency department (ED) setting, as it expedites expert assessments for psychiatric patients. However, limited knowledge exits for TP’s effectiveness in the ED setting, as well as the process of implementing TP in this setting.

OBJECTIVE: This scoping review aimed to review the existing evidence for the administrative and clinical outcomes for TP in the ED setting and to identify the barriers and facilitators to implementing TP in this setting.

METHODS: The scoping review was conducted according to the guidelines for the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews). Three electronic databases were examined: PubMed, Embase, and Web of Science. The databases were searched from January 2013 to April 2023 for papers and their bibliography. A total of 2816 potentially relevant papers were retrieved from the initial search. Studies were screened and selected independently by 2 authors.

RESULTS: A total of 11 articles were included. Ten papers reported on administrative and clinical outcomes of TP use in the ED setting and 1 on the barriers and facilitators of its implementation. TP is used in urban and rural areas and for settings with and with no on-site psychiatric services. Evidence shows that TP reduced waiting time for psychiatric evaluation, but in some studies, it was associated with prolonged total length of stay in the ED compared with in-person evaluation. Findings indicate lower admission rates in patients assessed with TP in the ED. Limited data were reported for TP costs, its use for involuntary commitment evaluations, and its use for particular subgroups of patients (eg, those with a particular diagnosis). A single paper examined TP implementation process in the ED, which explored the barriers and facilitators for implementation among patients and staff in a rural setting.

CONCLUSIONS: Based on the extant studies, TP seems to be generally feasible and acceptable to key stakeholders. However, this review detected a gap in the literature regarding TP’s effectiveness and implementation process in the ED setting. Specific attention should be paid to the examination of this service for specific groups of patients, as well as its use to enable assessments for possible involuntary commitment.

PMID:39008831 | DOI:10.2196/51814

J Am Med Inform Assoc. 2024 Jul 15:ocae186. doi: 10.1093/jamia/ocae186. Online ahead of print.

ABSTRACT

OBJECTIVE: Returning aggregate study results is an important ethical responsibility to promote trust and inform decision making, but the practice of providing results to a lay audience is not widely adopted. Barriers include significant cost and time required to develop lay summaries and scarce infrastructure necessary for returning them to the public. Our study aims to generate, evaluate, and implement ChatGPT 4 lay summaries of scientific abstracts on a national clinical study recruitment platform, ResearchMatch, to facilitate timely and cost-effective return of study results at scale.

MATERIALS AND METHODS: We engineered prompts to summarize abstracts at a literacy level accessible to the public, prioritizing succinctness, clarity, and practical relevance. Researchers and volunteers assessed ChatGPT-generated lay summaries across five dimensions: accuracy, relevance, accessibility, transparency, and harmfulness. We used precision analysis and adaptive random sampling to determine the optimal number of summaries for evaluation, ensuring high statistical precision.

RESULTS: ChatGPT achieved 95.9% (95% CI, 92.1-97.9) accuracy and 96.2% (92.4-98.1) relevance across 192 summary sentences from 33 abstracts based on researcher review. 85.3% (69.9-93.6) of 34 volunteers perceived ChatGPT-generated summaries as more accessible and 73.5% (56.9-85.4) more transparent than the original abstract. None of the summaries were deemed harmful. We expanded ResearchMatch’s technical infrastructure to automatically generate and display lay summaries for over 750 published studies that resulted from the platform’s recruitment mechanism.

DISCUSSION AND CONCLUSION: Implementing AI-generated lay summaries on ResearchMatch demonstrates the potential of a scalable framework generalizable to broader platforms for enhancing research accessibility and transparency.

PMID:39008829 | DOI:10.1093/jamia/ocae186

Hum Reprod. 2024 Jul 15:deae167. doi: 10.1093/humrep/deae167. Online ahead of print.

ABSTRACT

STUDY QUESTION: Does luteal estradiol (E2) pretreatment give a similar number of retrieved oocytes compared to no-pretreatment in advanced-aged women stimulated with corifollitropin alfa in an antagonist protocol?

SUMMARY ANSWER: Programming antagonist cycles with luteal E2 gave similar number of retrieved oocytes compared to no-pretreatment in women aged 38-42 years.

WHAT IS KNOWN ALREADY: Programming antagonist cycles with luteal E2 pretreatment is a valuable tool to organize the IVF procedure better and is safe without any known impact on cycle outcome. However, variable effects were observed on the number of retrieved oocytes depending on the treated population. In advanced-age women, recruitable follicles tend to decrease in number and to be more heterogeneous in size but it remains unclear if estradiol pretreatment could change the oocyte yield through its negative feed-back effect on FSH intercycle rise.

STUDY DESIGN, SIZE, DURATION: This non-blinded randomized controlled non-inferiority trial was conducted between 2016 and 2022 with centrally computerized randomization and concealed allocation. Participants were 324 women aged 38-42 years undergoing IVF treatment. The primary endpoint was the total number of retrieved oocytes. Statistical analysis was performed with one-sided alpha risk of 2.5% and 95% confidence interval (CI) with the non-inferiority of E2 pretreatment proved by a P value <0.025 and a lower delta margin of the CI within two oocytes compared to no pretreatment. Secondary endpoints were duration and total dosage of recombinant FSH, cancellation rate, percentage of oocyte pick-up (OPU) on working days, total number of metaphase II oocytes and obtained embryos, fresh transfer live birth rate, and cumulative live birth rate.

PARTICIPANTS/MATERIALS, SETTING, METHODS: This multicentric study enrolled women with regular cycles, weight >50 kg and body mass index <32, IVF cycle 1-2. According to randomization, micronized estradiol 2 mg twice a day was started on days 20-24 and continued until Wednesday beyond the onset of menses followed by administration of corifollitropin alfa on Friday, i.e. stimulation (S)1 or from D1-3 of a natural cycle in unpretreated patients. GnRH antagonist was started at S6 and additional FSH at S8.

MAIN RESULTS AND THE ROLE OF CHANCE: Basal characteristics were similar in patients randomized in E2 pretreated (n = 164) and non-pretreated (n = 160) groups (intended to treat (ITT) population). A total of 291 patients started treatment (per protocol (PP) population), 147 in E2 pretreated group with a mean number [SD] of pre-treatment days 9.8 [2.6] and 144 in the non-pretreated group. Despite advanced age, oocyte yields ranged from 0 to 29 in both groups with a median number of 6 retrieved oocytes in accordance with a mean anti-Müllerian hormone (AMH) level above 1.2 ng/ml. We demonstrated the non-inferiority of E2 pretreatment with a mean difference of -0.1 oocyte 95% CI [-1.5; 1.3] P = 0.004 in the PP population and a mean difference of -0.44 oocyte [-1.84; 0.97] P = 0.014 in the ITT population. Oocyte retrieval was more often on working days in E2 pretreated patients (91.9 versus 74.2%, P < 0.001). In patients reaching OPU, the duration of stimulation was statistically significantly longer (11.7 [1.7] versus 10.8 [1.8] days, P < 0.001) and the extra FSH dosage in addition to corifollitropin alfa was statistically significantly higher (1040 [548] versus 778 [504] IU, P < 0.001) in E2 pretreated than non-pretreated patients. We did not observe any significant differences in the number of retrieved oocytes (8.4 [6.1] versus 9.1 [6.0]), in the number of Metaphase 2 oocytes (7 [5.5] versus 7.3 [5.2]) nor in the number of obtained embryos (5 [4.6] versus 5.2 [4.2]) in E2 pretreated patients compared to non-pretreated patients. The live birth rate after fresh transfer (16.2% versus 18.5%, respectively), and the cumulative live birth rate per patient (17.7% versus 22.9%, respectively) were similar in both groups. Among the PP population, 31.6% of patients fulfilled the criteria for group 4 of Poseïdon classification (AMH <1.2 ng/ml and/or antral follicle count <5). In this sub-group of patients, we observed in contrast a statistically higher number of retrieved oocytes in E2 pretreated patients compared to non-pretreated (5.1 [3.8] versus 3.4 [2.7], respectively, the mean difference of +1.7 oocyte [0.2; 3.2] P = 0.022) but without significant difference in the cumulative live birth rate per patient (15.7% versus 7.3%, respectively).

LIMITATIONS, REASONS FOR CAUTION: Our stimulated women older than 38 years obtained a wide range of collected oocytes suggesting very different stages of ovarian aging in both groups. E2 pretreatment is more likely to increase oocyte yield at the stage of ovarian aging characterized by asynchrony of a reduced follicular cohort. Another limitation is the sample size in sub-group analysis of patients with AMH <1.2 ng/ml. Finally, the absence of placebo for pretreatment could also introduce possible bias.

WIDER IMPLICATIONS OF THE FINDINGS: Programming antagonist cycles with luteal E2 pretreatment seems a useful tool in advanced age women to better schedule oocyte retrievals on working days. However, the potential benefit of the number of collected oocytes remains to be demonstrated in a larger population displaying the characteristics of decreased ovarian reserve encountered in Poseïdon classification.

STUDY FUNDING/COMPETING INTEREST(S): Research grant from (MSD) Organon, France. I.C., S.D., B.B., X.M., S.G., and C.J. have no conflict of interest with this study. I.C.D. declares fees as speaker from Merck KGaA, Gedeon Richter, MSD (Organon, France), Ferring, Theramex, and IBSA and participation on advisory board from Merck KGaA. I.C.D. also declares consulting fees, and travel and meeting support from Merck KGaA. N.M. declares grants paid to their institution from MSD (Organon, France); consulting fees from MSD (Organon, France), Ferring, and Merck KGaA; honoraria from Merck KGaA, General Electrics, Genevrier (IBSA Pharma), and Theramex; support for travel and meetings from Theramex, Merck KGaG, and Gedeon Richter; and equipment paid to their institution from Goodlife Pharma. N.C. declares grants from IBSA Pharma, Merck KGaA, Ferring, and Gedeon Richter; support for travel and meetings from IBSA Pharma, Merck KGaG, MSD (Organon, France), Gedeon Richter, and Theramex; and participation on advisory board from Merck KGaA. A.G.L. declares fees as speaker from Merck KGaA, Gedeon Richter, MSD (Organon, France), Ferring, Theramex, and IBSA.

TRIAL REGISTRATION NUMBER: ClinicalTrials.gov NCT02884245.

TRIAL REGISTRATION DATE: 29 August 2016.

DATE OF FIRST PATIENT’S ENROLMENT: 4 November 2016.

PMID:39008826 | DOI:10.1093/humrep/deae167

Hum Reprod. 2024 Jul 15:deae161. doi: 10.1093/humrep/deae161. Online ahead of print.

ABSTRACT

STUDY QUESTION: Is there an elevated risk of cyanotic congenital heart defects (CCHD) among livebirths following infertility treatments?

SUMMARY ANSWER: In this population-based study of single livebirths, infertility treatment (either ART or non-ART) was associated with a higher prevalence of CCHD among livebirths.

WHAT IS KNOWN ALREADY: The use of infertility treatment has been on the rise over the past few decades. However, there are limited studies assessing the risk of major cardiac defects following infertility treatments.

STUDY DESIGN, SIZE, DURATION: A retrospective cohort study of livebirth data from the National Vital Statistics System (NVSS) was conducted, comprising of 9.6 million singleton livebirths among first-time mothers aged 15-49 years from 2016 to 2022.

PARTICIPANTS/MATERIALS, SETTING, METHODS: Information on infertility treatment use and CCHD was obtained from the health and medical information section of birth certificates, which was completed by healthcare staff after reviewing medical records. Logistic regression models were used to estimate odds ratios (OR) and 95% CI. Entropy balancing weighting analysis and probabilistic bias analysis were also performed.

MAIN RESULTS AND THE ROLE OF CHANCE: The proportion of births following infertility treatment increased from 1.9% (27 116) to 3.1% (43 510) during the study period. Overall, there were 5287 cases of CCHD resulting in a prevalence of 0.6 per 1000 livebirths. The prevalence was 1.2 per 1000 live births among infertility treatment users (ART: 1.1 per 1000 livebirths; non-ART: 1.3 per 1000 livebirths) while that for naturally conceived births was 0.5 per 1000 livebirths. Compared to naturally conceived births, the use of any infertility treatment (OR: 2.06, 95% CI: 1.82-2.33), either ART (OR: 2.02, 95% CI: 1.73-2.36) or other infertility treatments (OR: 2.12, 95% CI: 1.74-2.33), was associated with higher odds of CCHD after adjusting for maternal and paternal age, race and ethnicity, and education, as well as maternal nativity, marital status, source of payment, smoking status, and pre-pregnancy measures of BMI, hypertension and diabetes. This association did not differ by the type of infertility treatment (ART versus other infertility treatments) (OR: 1.04, 95% CI: 0.82-1.33, P = 0.712), and was robust to the presence of exposure and outcome misclassification bias and residual confounding.

LIMITATIONS, REASONS FOR CAUTION: The findings are only limited to livebirths. We did not have the capacity to examine termination data, but differential termination by mode of conception has not been supported by previous studies designed to consider it. Infertility treatment use was self-reported, leading to the potential for selection bias and misclassification for infertility treatment and CCHD. However, the association persisted when systematic bias as well as exposure and outcome misclassification bias were accounted for in the analyses. Information on the underlying etiology of infertility relating to either maternal, paternal, or both factors, data on specific types of ART and other infertility treatments, as well as information on subtypes of CCHD, were all not available.

WIDER IMPLICATIONS OF THE FINDINGS: In light of the increasing trend in the use of infertility treatment in the USA, and elsewhere, the finding of the current study holds significant importance for the clinical and public health of reproductive-aged individuals. The data show that the use of infertility treatment may expose offspring to elevated odds of severe congenital heart defects such as CCHD studied here. These findings cannot be interpreted causally. While our findings can assist in preconception counseling and prenatal care for pregnancies conceived by either ART or other infertility treatments, they also support some current recommendations that pregnancies resulting from infertility treatments undergo fetal echocardiography screening.

STUDY FUNDING/COMPETING INTEREST(S): No funding was sought for the study. The authors declare that they have no conflict of interest.

TRIAL REGISTRAION NUMBER: N/A.

PMID:39008825 | DOI:10.1093/humrep/deae161

Nutr Rev. 2024 Jul 15:nuae093. doi: 10.1093/nutrit/nuae093. Online ahead of print.

ABSTRACT

CONTEXT: Energy intake may differ across the menstrual cycle, with some studies identifying greater energy intake in the luteal phase (LP) compared with the follicular phase (FP) and others finding no clear differences. To date, no study has systematically synthesized the available data to draw more definite conclusions while considering any methodological inconsistencies between studies.

OBJECTIVE: The aim was to conduct a systematic review/meta-analysis in an effort to determine if there are differences in energy intake between the FP and LP.

DATA SOURCES: A systematic search strategy was developed and the search was conducted in 5 databases for studies that investigated any changes in energy intake across menstrual phases.

DATA EXTRACTION: Using Covidence, studies were identified and included if they contained individuals between the ages of 18 and 45 years, maintained an average body mass index (BMI) of 18.5-25 kg/m2, had no history of disordered eating, and included energy intake and menstrual cycle measurements in the FP and LP.

DATA ANALYSIS: Effect sizes were calculated for each study and a random-effects model was used to pool the results of each study.

RESULTS: Fifteen datasets were included consisting of 330 female participants with a mean age of 26 ± 4 years and mean BMI of 22.4 ± 2.3 kg/m2. Overall, there was a statistically significant difference (standardized mean difference = 0.69; P = .039) with increased energy intake in the LP compared with the FP (crude 168 kcal⋅d-1 average difference between phases).

CONCLUSION: Energy intake was found to be greater in the LP compared with the FP, providing insight into the effect of the menstrual cycle on energy intake. However, there were repeated methodological inconsistencies and future work should strive to utilize best practices for both energy intake measurement and menstrual phase specification.

PMID:39008822 | DOI:10.1093/nutrit/nuae093

Neurology. 2024 Aug 13;103(3):e209622. doi: 10.1212/WNL.0000000000209622. Epub 2024 Jul 15.

ABSTRACT

BACKGROUND AND OBJECTIVES: Persons with epilepsy (PwE) have a higher risk of developing psychiatric comorbidities compared with the general population. There is limited knowledge about the prevalence of multiple psychiatric conditions in PwE. We summarize the current evidence on the prevalence of multipsychiatric comorbidities in PwE compared with persons without epilepsy.

METHODS: A systematic review of multipsychiatric comorbidities in PwE compared with persons without epilepsy was performed, and the results were reported using the Preferred Reporting Items of Systematic Reviews and Meta-analyses reporting standards. The search was conducted from January 1945 to June 2023 in Ovid MEDLINE. Embase, and PsycINFO, using the search terms related to “epilepsy,” “psychiatric comorbidity,” and “multimorbidity,” combined with psychiatric disorders. Abstracts were reviewed in duplicate, and data were independently extracted using standard proforma. Data describing multipsychiatric comorbidities in PwE compared with persons without epilepsy were recorded. Descriptive statistics and, when feasible, meta-analyses are presented. The risk of bias of the studies was assessed using the Newcastle-Ottawa Scale and the International League Against Epilepsy tool.

RESULTS: A total of 12,841 records were identified from the systematic database search, and 15 studies met the eligibility criteria. All included studies were deemed high-quality in risk of bias according to both tools. The prevalence of multipsychiatric comorbidity was greater in persons with compared with those without epilepsy. The pooled prevalence of concomitant depression and anxiety disorder in PwE in 2 population-based studies was 15 of 163 (9.2%), which was significantly higher than 250 of 10,551 (2.4%) in patients without epilepsy (odds ratio [OR] 3.7, 95% CI 2.1-6.5, p-value <0.001, I² = 0%, Cochran Q p-value for heterogeneity = 0.84). In 2 hospital-based studies, the prevalence of concomitant depression and attention-deficit/hyperactivity disorder in PwE (14/97, 14.4%) was significantly higher than in patients without epilepsy (5/126, 3.9%), with an OR 5.2 (95% CI 1.8-15.0, p-value = 0.002, I² = 0%, Cochran Q p-value for heterogeneity = 0.79).

DISCUSSION: PwE experience elevated levels of multipsychiatric comorbidity compared with those without epilepsy. However, very few studies have empirically evaluated the extent of multipsychiatric comorbidity in PwE compared with persons without epilepsy nor their associations and consequences to prognosis in PwE.

PMID:39008805 | DOI:10.1212/WNL.0000000000209622

Neurology. 2024 Aug 13;103(3):e209657. doi: 10.1212/WNL.0000000000209657. Epub 2024 Jul 15.

ABSTRACT

BACKGROUND AND OBJECTIVES: While emerging theories suggest that vascular dysfunction may occur concurrently with the amyloid cascade in Alzheimer disease (AD) pathogenesis, the role of vascular components as primary neurodegeneration triggers remains uncertain. The aim of this retrospective, population-based cohort study conducted in Korea was to explore the link between nonarteritic anterior ischemic optic neuropathy (NAION) and dementia risk.

METHODS: In this nationwide, population-based, retrospective cohort study, we identified newly diagnosed NAION from 2010 to 2017 in the Korean National Health Insurance Service database. The primary outcome was new dementia diagnoses confirmed by new ICD-10 claims coupled with antidementia medication prescriptions. We assessed dementia risk using hazard ratios (HRs) with 95% CIs over an average 2.69-year follow-up after a 1-year lag period.

RESULTS: The cohort consisted of 42,943 patients with NAION and 214,715 age-matched and sex-matched controls without NAION (mean age 61.37 years ± 10.75 SD, 55.48% female). The study found a higher risk of all-cause dementia (ACD; HR 1.28, 95% CI 1.20-1.36), AD (HR 1.27, 95% CI 1.18-1.36), vascular dementia (VaD; HR 1.31, 95% CI 1.09-1.58), and other dementia (HR 1.39, 95% CI 1.11-1.73) among patients with NAION, regardless of other potential confounding factors such as age, sex, lifestyle behaviors, economic status, and preexisting health conditions. In subgroup analysis, the associations between NAION and ACD were stronger in the younger age group (HR 1.83 for those younger than 65 years vs 1.23 for those 65 years or older; p for interaction <0.001). Moreover, the association of NAION with both ACD and VaD was particularly strong among current smokers.

DISCUSSION: We found a significant association between NAION and increased risk for ACD, AD, VaD, and other dementia even after adjusting for potential confounders such as lifestyle, health conditions, and demographic factors within a nationwide cohort. This study highlights the potential role of vascular pathology in dementia progression and suggests that NAION may serve as a robust predictor for dementia, highlighting the need for comprehensive neurologic assessment in patients with NAION. Further research is needed to clarify the association between NAION and dementia risk.

PMID:39008797 | DOI:10.1212/WNL.0000000000209657

Med Phys. 2024 Jul 15. doi: 10.1002/mp.17291. Online ahead of print.

ABSTRACT

BACKGROUND: Vessel-wall volume and localized three-dimensional ultrasound (3DUS) metrics are sensitive to the change of carotid atherosclerosis in response to medical/dietary interventions. Manual segmentation of the media-adventitia boundary (MAB) and lumen-intima boundary (LIB) required to obtain these metrics is time-consuming and prone to observer variability. Although supervised deep-learning segmentation models have been proposed, training of these models requires a sizeable manually segmented training set, making larger clinical studies prohibitive.

PURPOSE: We aim to develop a method to optimize pre-trained segmentation models without requiring manual segmentation to supervise the fine-tuning process.

METHODS: We developed an adversarial framework called the unsupervised shape-and-texture generative adversarial network (USTGAN) to fine-tune a convolutional neural network (CNN) pre-trained on a source dataset for accurate segmentation of a target dataset. The network integrates a novel texture-based discriminator with a shape-based discriminator, which together provide feedback for the CNN to segment the target images in a similar way as the source images. The texture-based discriminator increases the accuracy of the CNN in locating the artery, thereby lowering the number of failed segmentations. Failed segmentation was further reduced by a self-checking mechanism to flag longitudinal discontinuity of the artery and by self-correction strategies involving surface interpolation followed by a case-specific tuning of the CNN. The U-Net was pre-trained by the source dataset involving 224 3DUS volumes with 136, 44, and 44 volumes in the training, validation and testing sets. The training of USTGAN involved the same training group of 136 volumes in the source dataset and 533 volumes in the target dataset. No segmented boundaries for the target cohort were available for training USTGAN. The validation and testing of USTGAN involved 118 and 104 volumes from the target cohort, respectively. The segmentation accuracy was quantified by Dice Similarity Coefficient (DSC), and incorrect localization rate (ILR). Tukey’s Honestly Significant Difference multiple comparison test was employed to quantify the difference of DSCs between models and settings, where p ≤ 0.05 $p,le ,0.05$ was considered statistically significant.

RESULTS: USTGAN attained a DSC of 85.7 ± 13.0 $85.7,pm ,13.0$ % in LIB and 86.2 ± 10.6 ${86.2},pm ,{10.6}$ % in MAB, improving from the baseline performance of 74.6 ± 30.7 ${74.6},pm ,{30.7}$ % in LIB (p < 10 – 12 $&lt;10^{-12}$ ) and 75.7 ± 28.9 ${75.7},pm ,{28.9}$ % in MAB (p < 10 – 12 $&lt;10^{-12}$ ). Our approach outperformed six state-of-the-art domain-adaptation models (MAB: p ≤ 3.63 × 10 – 7 $p le 3.63,times ,10^{-7}$ , LIB: p ≤ 9.34 × 10 – 8 $p,le ,9.34,times ,10^{-8}$ ). The proposed USTGAN also had the lowest ILR among the methods compared (LIB: 2.5%, MAB: 1.7%).

CONCLUSION: Our framework improves segmentation generalizability, thereby facilitating efficient carotid disease monitoring in multicenter trials and in clinics with less expertise in 3DUS imaging.

PMID:39008794 | DOI:10.1002/mp.17291