ABSTRACT
Transfer learning plays a pivotal role in addressing the paucity of data, expediting training processes, and enhancing model performance. Nonetheless, the prevailing practice of transfer learning predominantly relies on pre-trained models designed for the natural image domain, which may not be well-suited for the medical image domain in grayscale. Recognizing the significance of leveraging transfer learning in medical research, we undertook the construction of class-balanced pediatric radiograph datasets collectively referred to as PedXnets, grounded in radiographic views using the pediatric radiographs collected over 24 years at Asan Medical Center. For PedXnets pre-training, approximately 70,000 X-ray images were utilized. Three different pre-training weights of PedXnet were constructed using Inception V3 for various radiation perspective classifications: Model-PedXnet-7C, Model-PedXnet-30C, and Model-PedXnet-68C. We validated the transferability and positive effects of transfer learning of PedXnets through pediatric downstream tasks including fracture classification and bone age assessment (BAA). The evaluation of transfer learning effects through classification and regression metrics showed superior performance of Model-PedXnets in quantitative assessments. Additionally, visual analyses confirmed that the Model-PedXnets were more focused on meaningful regions of interest.
Subject(s)
Deep Learning , Fractures, Bone , Humans , Child , Machine Learning , RadiographyABSTRACT
The emergence of Chat Generative Pre-trained Transformer (ChatGPT), a chatbot developed by OpenAI, has garnered interest in the application of generative artificial intelligence (AI) models in the medical field. This review summarizes different generative AI models and their potential applications in the field of medicine and explores the evolving landscape of Generative Adversarial Networks and diffusion models since the introduction of generative AI models. These models have made valuable contributions to the field of radiology. Furthermore, this review also explores the significance of synthetic data in addressing privacy concerns and augmenting data diversity and quality within the medical domain, in addition to emphasizing the role of inversion in the investigation of generative models and outlining an approach to replicate this process. We provide an overview of Large Language Models, such as GPTs and bidirectional encoder representations (BERTs), that focus on prominent representatives and discuss recent initiatives involving language-vision models in radiology, including innovative large language and vision assistant for biomedicine (LLaVa-Med), to illustrate their practical application. This comprehensive review offers insights into the wide-ranging applications of generative AI models in clinical research and emphasizes their transformative potential.
Subject(s)
Artificial Intelligence , Radiology , Humans , Diagnostic Imaging , Software , LanguageABSTRACT
Recent advances in contrastive learning have significantly improved the performance of deep learning models. In contrastive learning of medical images, dealing with positive representation is sometimes difficult because some strong augmentation techniques can disrupt contrastive learning owing to the subtle differences between other standardized CXRs compared to augmented positive pairs; therefore, additional efforts are required. In this study, we propose intermediate feature approximation (IFA) loss, which improves the performance of contrastive convolutional neural networks by focusing more on positive representations of CXRs without additional augmentations. The IFA loss encourages the feature maps of a query image and its positive pair to resemble each other by maximizing the cosine similarity between the intermediate feature outputs of the original data and the positive pairs. Therefore, we used the InfoNCE loss, which is commonly used loss to address negative representations, and the IFA loss, which addresses positive representations, together to improve the contrastive network. We evaluated the performance of the network using various downstream tasks, including classification, object detection, and a generative adversarial network (GAN) inversion task. The downstream task results demonstrated that IFA loss can improve the performance of effectively overcoming data imbalance and data scarcity; furthermore, it can serve as a perceptual loss encoder for GAN inversion. In addition, we have made our model publicly available to facilitate access and encourage further research and collaboration in the field.
Subject(s)
Deep Learning , Neural Networks, Computer , Humans , Radiography, ThoracicABSTRACT
OBJECTIVE: To determine whether fragment removal on in vitro fertilization (IVF) day 2 improved the subsequent development and pregnancy outcomes of fragmented embryos compared to similar-grade embryos without fragment removal. METHODS: This study was a retrospective analysis involving 191 IVF cycles in which all embryos had over 10% fragmentation (grade 3 or 4) on day 2 of the IVF-embryo transfer cycle from March 2015 to December 2017. IVF cycles were divided into the fragment removal group (n=87) and the no fragment removal group (n=104) as a control cohort. Before fragment removal, embryos with fragmentation on day 2 were incubated in Ca2+- and Mg2+-free biopsy medium under paraffin oil for 30 minutes. Microsurgical fragment removal was performed with later-assisted hatching and a handmade suction micropipette that had an outer diameter of 30 µm. RESULTS: There were no significant differences in the characteristics of the patients between the control and the fragment removal groups. After fragment removal and subsequent in vitro culture for 24 hours, the number of blastomeres (7.1±1.7 vs. 6.9±1.6) was comparable between the transferred embryos in the two groups, but the morphological grade of the embryos in the fragment removal group (1.9±0.7) was significantly higher than that of the control group (3.1±0.5, p < 0.01). The clinical pregnancy (43.7%) and implantation rates (25.8%) in the fragment removal group were significantly higher than those in the control group (28.8% and 14.0%, respectively; p < 0.05). CONCLUSION: Early fragment removal on day 2 significantly improved the subsequent development and pregnancy outcomes of fragmented embryos.
Subject(s)
Female , Humans , Pregnancy , Biopsy , Blastomeres , Cohort Studies , Embryonic Structures , Fertilization in Vitro , In Vitro Techniques , Paraffin , Pregnancy Outcome , Retrospective Studies , SuctionABSTRACT
OBJECTIVE: Correlations between semen parameters and sperm DNA fragmentation index (DFI) were investigated to identify characteristics of sperm without DNA damage that could be used in selecting sperm for intracytoplasmic sperm injection (ICSI). Pregnancy outcomes were compared to determine whether in vitro fertilization (IVF) or ICSI is a better choice for patients who have sperm with a high-DFI. METHODS: Semen analysis was carried out in 388 patients who visited our IVF center for the first time to investigate correlations between sperm DFI and semen parameters. In addition, 1,102 IVF cycles in 867 patients were carried out in the present study; 921 cycles in the low-DFI group (DFI <30%) and 181 cycles in the high-DFI group (DFI ≥30%). Both the low- and high-DFI groups were subdivided into IVF and ICSI cycle groups. RESULTS: Sperm DFI showed significant inverse correlations with sperm motility (r=−0.435, p<0.001) and morphology (r=−0.153, p<0.05). Sperm DFI also showed significant correlations with rapid motility (r=−0.436, p<0.001), and the kinetic parameters of average-path velocity (r=−0.403) and linearity (r=−0.412). Although there was no significant difference in the pregnancy rates between IVF (48.6%) and ICSI (44.8%) in the low-DFI group, the pregnancy rate of ICSI cycles (44.8%, p<0.05) was significantly higher than IVF cycles (25.0%) in the high-DFI group. No significant difference was observed in the abortion rates between the low-DFI (52 of 921, 5.6%) and high-DFI groups (7 of 181, 3.8%). CONCLUSION: ICSI is a better choice than IVF for improving the pregnancy outcomes of patients who have sperm with a high DFI.
Subject(s)
Female , Humans , Pregnancy , Abortion, Induced , DNA Damage , DNA Fragmentation , DNA , Fertilization in Vitro , Pregnancy Outcome , Pregnancy Rate , Semen , Semen Analysis , Sperm Injections, Intracytoplasmic , Sperm Motility , SpermatozoaABSTRACT
OBJECTIVE: To investigate psychological distress and fertility quality of life (FertiQoL) in infertile Korean women, and to investigate whether a correlation exists between psychological distress and FertiQoL. METHODS: Participants in this study were made up of 141 infertile women and 65 fertile women. We conducted a survey on psychological distress (using the Depression Anxiety Stress Scales [DASS]-42 questionnaire) and administered a FertiQoL questionnaire. The levels of stress hormones (adrenocorticotropic hormone [ACTH] and cortisol) in serum were assessed. RESULTS: The scores for depression (13.7±8.4), anxiety (10.7±6.4), and stress (18.0±8.3) among the infertile women were significantly higher than the scores for depression (9.4±7.5), anxiety (6.6±6.0), and stress (12.2±8.3, p<0.001) among the fertile women. There was no difference in the scores for depression (13.5±8.2, 13.8±8.6), anxiety (10.0±6.2, 11.5±6.6) and stress (17.7±8.4, 18.4±8.1) between younger (≤34) and older (≥35) participants. The mind-body (r =–0.495) and emotional (r =–0.590) subscales showed a higher negative correlation with stress compared with other scales of psychological distress. At the same time, the social (r =–0.537) and relational (r =–0.385) subscales showed a higher negative correlation with depression. Levels of cortisol and ACTH in infertile women were 9.1 µg/mL and 11.9 pg/mL, respectively, which are within normal ranges. CONCLUSION: The levels of psychological distress and quality of life in infertile Korean women seem to require psychological intervention. This study provides a baseline measurement of psychological distress and FertiQoL in infertile women in Korea, which will be available for developing psychological interventions for infertile Korean women.
Subject(s)
Female , Humans , Adrenocorticotropic Hormone , Anxiety , Depression , Fertility , Hydrocortisone , Korea , Quality of Life , Reference Values , Weights and MeasuresABSTRACT
OBJECTIVE: This study was carried out to investigate the correlations of the sperm DNA fragmentation index (DFI) with semen parameters and apoptosis, and to investigate the effects of density-gradient centrifugation (DGC) and magnetic-activated cell sorting (MACS) on reducing the proportion of sperm with DNA fragmentation and protamine deficiency. METHODS: Semen analysis and a sperm DNA fragmentation assay were performed to assess the correlations between semen parameters and the DFI in 458 semen samples. Sperm with progressive motility or non-apoptosis were isolated by DGC or MACS, respectively, in 29 normozoospermic semen samples. The effects of DGC or MACS alone and of DGC and MACS combined on reducing the amount of sperm in the sample with DNA fragmentation and protamine deficiency were investigated. RESULTS: The sperm DFI showed a significant correlation (r=–0.347, p<0.001) with sperm motility and morphology (r=–0.114, p<0.05) but not with other semen parameters. The DFI (11.5%±2.0%) of semen samples was significantly reduced by DGC (8.1%±4.1%) or MACS alone (7.4%±3.9%) (p<0.05). The DFI was significantly further reduced by a combination of DGC and MACS (4.1%±1.3%, p<0.05). Moreover, the combination of DGC and MACS (1.6%±1.1%, p<0.05) significantly reduced the protamine deficiency rate of semen samples compared to DGC (4.4%±3.2%) or MACS alone (3.4%±2.2%). CONCLUSION: The combination of DGC and MACS may be an effective method to isolate high-quality sperm with progressive motility, non-apoptosis, high DNA integrity, and low protamine deficiency in clinical use.
Subject(s)
Apoptosis , Centrifugation , Centrifugation, Density Gradient , Chromatin , DNA Fragmentation , DNA , Methods , Product Packaging , Semen , Semen Analysis , Sperm Motility , SpermatozoaABSTRACT
OBJECTIVE: The purpose of this study is to verify association between HPV viral load and recurrence after conization and analyse clinicopathological characteristics in patients with carcinoma in situ of cervix. METHODS: We reviewed the chart of 154 women with carcinoma in situ who underwent conization from January 2003 to December 2005. Pap smear and high risk HPV test was performed before treatment and colposcopic examniation with biopsy was performed in necessary. After operation, repeated Pap smear and quantitative high risk HPV test (Hybrid capture II test, HC II) were performed in three month after conization for recurrence of disease. We studied the medical records of the patients, analyzed demographic characteristics and conducted the correlation between the prognostic factors of the age, parity, presence of high risk HPV infection before and after the operation, existence of invasion of resection margin after conization and remaining or recurrent lesions. RESULTS: Positive of cone margin showed significantly high recurrence rate than negative cone margin (p<0.05). And posttreatment HPV viral load was significantly higher in recurrent patients (p<0.05). But initial Pap smear, age and parity has not shown in difference. CONCLUSION: We can conclude that a clear association between positive cone margin, high HPV viral load and recurrence of disease after conization. A patient with positive resection margin after or high viral load after conization should be evaluated and treated aggressively.
Subject(s)
Female , Humans , Biopsy , Carcinoma in Situ , Cervix Uteri , Conization , Medical Records , Parity , Recurrence , Viral LoadABSTRACT
OBJECTIVE: The purpose of this study is to verify the correlation between gestational sac volume (GSV) from automatic tracing mode the VOCAL-imaging program and gestational age and to construct the nomogram of gestational sac volume as the new biometric parameter in early pregnancy. METHODS: The cross-sectional study has been conducted in 242 uncomplicated singleton pregnancies. 47 cases were excluded due to early pregnancy failure, fetal malformations, elective abortion, age discrepancy, etc. In 195 uncomplicated singleton pregnancies from 5 to 12 weeks' menstrual age, gestational sac volume, mean sac diameter and crown-lump length were measured for the assessment of gestational age. Gestational sac volumetry was carried out with automatic method and manual method of the VOCALTM (Virtual Organ Computer- aided AnaLysis) for the comparison between two methods. The collected data were analyzed for mean, standard deviation, 90% reference interval, 5th, 50th and 95th percentiles of gestational sac volume, mean sac diameter and CRL, and the nomogram were constructed. RESULTS: Polynomial regression analysis demonstrated the statistically significant positive correlation between gestational age and gestational sac volume by automatic tracing mode (R2 0.826, p<0.001), gestational sac volume by manual tracing mode (R2 0.844, p<0.0001), mean sac diameter (R2 0.763, p<0.0001) and crown-lump length (R2 0.950, p<0.0001). The 5th, 50th and 95th percentiles of the gestational sac volume were calculated and the nomogram was tabulated. In determining gestational age, the standard deviation (SD) of gestational sac volume by automatic tracing mode is 5.6 days, the SD of gestational sac volume by manual tracing mode is 5.2 days and the SD of MSD is 6.6 days. CONCLUSION: we can conclude that three-dimensional GSV with automatic tracing mode of the VOCAL-imaging program can be used as the new biometric measurement in determining gestational age. Gestational sac volumetry with automatic tracing mode of the VOCAL-imaging program have been proven available and convenient method and it can be recommended in 5-7 weeks' of gestation, when CRL is not clearly visualized.
Subject(s)
Pregnancy , Cross-Sectional Studies , Gestational Age , Gestational Sac , NomogramsABSTRACT
A 75-year-old female presented with a 1-month history of low abd pain. A 15 x 13 cm solid mass was noted in pelvic cavity by imaging study. The patient underwent an exploratory laparotomy, total abdominal hysterectomy, bilateral salpingo-oophorectomy, pelvic lymph node dissection and omentectomy. Immunohistochemistry aided in the final diagnosis of a malignant peripheral nerve sheath tumor. The patient had several poor prognosis factor and relapsed into tumor and died 5 month after surgery. An extremely rare case of malignant peripheral nerve sheath tumor of pelvic cavity is reported with a brief review of the literature.