RESUMEN
In medicine, confounding variables in a generalized linear model are often adjusted; however, these variables have not yet been exploited in a non-linear deep learning model. Sex plays important role in bone age estimation, and non-linear deep learning model reported their performances comparable to human experts. Therefore, we investigate the properties of using confounding variables in a non-linear deep learning model for bone age estimation in pediatric hand X-rays. The RSNA Pediatric Bone Age Challenge (2017) dataset is used to train deep learning models. The RSNA test dataset is used for internal validation, and 227 pediatric hand X-ray images with bone age, chronological age, and sex information from Asan Medical Center (AMC) for external validation. U-Net based autoencoder, U-Net multi-task learning (MTL), and auxiliary-accelerated MTL (AA-MTL) models are chosen. Bone age estimations adjusted by input, output prediction, and without adjusting the confounding variables are compared. Additionally, ablation studies for model size, auxiliary task hierarchy, and multiple tasks are conducted. Correlation and Bland-Altman plots between ground truth and model-predicted bone ages are evaluated. Averaged saliency maps based on image registration are superimposed on representative images according to puberty stage. In the RSNA test dataset, adjusting by input shows the best performances regardless of model size, with mean average errors (MAEs) of 5.740, 5.478, and 5.434 months for the U-Net backbone, U-Net MTL, and AA-MTL models, respectively. However, in the AMC dataset, the AA-MTL model that adjusts the confounding variable by prediction shows the best performance with an MAE of 8.190 months, whereas the other models show the best performances by adjusting the confounding variables by input. Ablation studies of task hierarchy reveal no significant differences in the results of the RSNA dataset. However, predicting the confounding variable in the second encoder layer and estimating bone age in the bottleneck layer shows the best performance in the AMC dataset. Ablations studies of multiple tasks reveal that leveraging confounding variables plays an important role regardless of multiple tasks. To estimate bone age in pediatric X-rays, the clinical setting and balance between model size, task hierarchy, and confounding adjustment method play important roles in performance and generalizability; therefore, proper adjusting methods of confounding variables to train deep learning-based models are required for improved models.
Asunto(s)
Aprendizaje Profundo , Radiología , Humanos , Niño , Rayos X , Factores de Confusión Epidemiológicos , RadiografíaRESUMEN
Training deep learning models on medical images heavily depends on experts' expensive and laborious manual labels. In addition, these images, labels, and even models themselves are not widely publicly accessible and suffer from various kinds of bias and imbalances. In this paper, chest X-ray pre-trained model via self-supervised contrastive learning (CheSS) was proposed to learn models with various representations in chest radiographs (CXRs). Our contribution is a publicly accessible pretrained model trained with a 4.8-M CXR dataset using self-supervised learning with a contrastive learning and its validation with various kinds of downstream tasks including classification on the 6-class diseases in internal dataset, diseases classification in CheXpert, bone suppression, and nodule generation. When compared to a scratch model, on the 6-class classification test dataset, we achieved 28.5% increase in accuracy. On the CheXpert dataset, we achieved 1.3% increase in mean area under the receiver operating characteristic curve on the full dataset and 11.4% increase only using 1% data in stress test manner. On bone suppression with perceptual loss, we achieved improvement in peak signal to noise ratio from 34.99 to 37.77, structural similarity index measure from 0.976 to 0.977, and root-square-mean error from 4.410 to 3.301 when compared to ImageNet pretrained model. Finally, on nodule generation, we achieved improvement in Fréchet inception distance from 24.06 to 17.07. Our study showed the decent transferability of CheSS weights. CheSS weights can help researchers overcome data imbalance, data shortage, and inaccessibility of medical image datasets. CheSS weight is available at https://github.com/mi2rl/CheSS .
Asunto(s)
Rayos X , Humanos , Curva ROC , Radiografía , Relación Señal-RuidoRESUMEN
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.
Asunto(s)
Aprendizaje Profundo , Redes Neurales de la Computación , Humanos , Radiografía TorácicaRESUMEN
Hematoxylin and eosin (H&E) staining is the gold standard modality for diagnosis in medicine. However, the dosage ratio of hematoxylin to eosin in H&E staining has not been standardized yet. Additionally, H&E stains fade out at various speeds. Therefore, the staining quality could differ among each image, and stain normalization is a critical preprocessing approach for training deep learning (DL) models, especially in long-term and/or multicenter digital pathology studies. However, conventional methods for stain normalization have some significant drawbacks, such as collapsing in the structure and/or texture of tissue. In addition, conventional methods must require a reference patch or slide. Meanwhile, DL-based methods have a risk of overfitting and/or grid artifacts. We developed a score-based diffusion model of colorization for stain normalization. However, mistransfer, in which the model confuses hematoxylin with eosin, can occur using a score-based diffusion model due to its high diversity nature. To overcome this mistransfer, we propose a stain separation method using sparse non-negative matrix factorization (SNMF), which can decompose pathology slide into Hematoxylin and Eosin to normalize each stain component. Furthermore, inpainting with overlapped moving window patches was used to prevent grid artifacts of whole slide image normalization. Our method can normalize the whole slide pathology images through this stain normalization pipeline with decent performance.
Asunto(s)
Algoritmos , Colorantes , Colorantes/química , Hematoxilina , Eosina Amarillenta-(YS) , Coloración y EtiquetadoRESUMEN
Artificial intelligence (AI) in radiology is a rapidly developing field with several prospective clinical studies demonstrating its benefits in clinical practice. In 2022, the Korean Society of Radiology held a forum to discuss the challenges and drawbacks in AI development and implementation. Various barriers hinder the successful application and widespread adoption of AI in radiology, such as limited annotated data, data privacy and security, data heterogeneity, imbalanced data, model interpretability, overfitting, and integration with clinical workflows. In this review, some of the various possible solutions to these challenges are presented and discussed; these include training with longitudinal and multimodal datasets, dense training with multitask learning and multimodal learning, self-supervised contrastive learning, various image modifications and syntheses using generative models, explainable AI, causal learning, federated learning with large data models, and digital twins.
Asunto(s)
Inteligencia Artificial , Radiología , Humanos , Estudios Prospectivos , Radiología/métodos , Aprendizaje Automático SupervisadoRESUMEN
A major responsibility of radiologists in routine clinical practice is to read follow-up chest radiographs (CXRs) to identify changes in a patient's condition. Diagnosing meaningful changes in follow-up CXRs is challenging because radiologists must differentiate disease changes from natural or benign variations. Here, we suggest using a multi-task Siamese convolutional vision transformer (MuSiC-ViT) with an anatomy-matching module (AMM) to mimic the radiologist's cognitive process for differentiating baseline change from no-change. MuSiC-ViT uses the convolutional neural networks (CNNs) meet vision transformers model that combines CNN and transformer architecture. It has three major components: a Siamese network architecture, an AMM, and multi-task learning. Because the input is a pair of CXRs, a Siamese network was adopted for the encoder. The AMM is an attention module that focuses on related regions in the CXR pairs. To mimic a radiologist's cognitive process, MuSiC-ViT was trained using multi-task learning, normal/abnormal and change/no-change classification, and anatomy-matching. Among 406 K CXRs studied, 88 K change and 115 K no-change pairs were acquired for the training dataset. The internal validation dataset consisted of 1,620 pairs. To demonstrate the robustness of MuSiC-ViT, we verified the results with two other validation datasets. MuSiC-ViT respectively achieved accuracies and area under the receiver operating characteristic curves of 0.728 and 0.797 on the internal validation dataset, 0.614 and 0.784 on the first external validation dataset, and 0.745 and 0.858 on a second temporally separated validation dataset. All code is available at https://github.com/chokyungjin/MuSiC-ViT.
Asunto(s)
Música , Humanos , Estudios de Seguimiento , Aprendizaje , Redes Neurales de la Computación , Curva ROCRESUMEN
OBJECTIVE: This study aimed to investigate the relationship between long working hours and chronic kidney disease (CKD) according to diabetic status. METHODS: Twelve thousand seven hundred three full-time employees without diabetes and 2136 with diabetes were included in this study. Participants were grouped according to working hours: ≤40, 41 to 52, and >52âh/week. Multiple logistic regression was used to evaluate the association between working hours and CKD prevalence. RESULTS: Participants with diabetes who worked 41 to 52âh/week showed 1.85 times higher odds of CKD (95% CI 1.15-2.96; Pâ=â0.0112) compared with those who worked ≤40âh/week after adjusting for covariates. An interaction between diabetes and long working hours was observed (P for interactionâ=â0.0212) in the model. CONCLUSION: Long working hours are associated with CKD in participants with diabetes. An interaction between long working hours and diabetes leading to CKD development may exist.
Asunto(s)
Diabetes Mellitus , Insuficiencia Renal Crónica , Diabetes Mellitus/epidemiología , Empleo , Femenino , Humanos , Masculino , Encuestas Nutricionales , Insuficiencia Renal Crónica/epidemiología , República de Corea/epidemiologíaRESUMEN
BACKGROUND: The use of topical antibiotics (TA) for prophylactic purposes after clean dermatologic procedures (CDP) is generally not recommended, and the prescription of TA needs to be individualized in consideration of each patient's situation and underlying disease. The aim of this study was to determine the proportion of patients who underwent CDP in outpatient settings and were prescribed TA inappropriately, as well as the factors that may affect the prescription of TA. METHODS: Outpatient visits coded for CDP were selected using claims data from the Health Insurance Review and Assessment Service in 2018. Of these, patients receiving TA prescriptions were classified as having inappropriate TA use, and the proportion was estimated through technical analysis. A logistic regression analysis was used to identify factors influencing inappropriate prescriptions. RESULTS: Data were analyzed using 423,651 visits, and TA was prescribed for approximately 1.9% of the visits. TA usage was higher among women (2.0%), 0-19 years of age (2.2%), medical aid (2.2%), clinic settings (2.4%), and metropolitan areas (2.0%). TA was prescribed more frequently in urology (8.6%), pediatrics (5.0%), and dermatology (4.2%) than in other specialties. CONCLUSION: The prescription rate of TA after CDP was 1.9% using the 1.4 million patient sample from the national health insurance claims data in Korea, which is equally weighted to represent 50 million people. Although the proportion of inappropriate TA prescriptions in Korea is lower than that in other nations, it cannot be overlooked because of the large number of cases. Efforts to improve quality are required to reduce the number of inappropriate prescriptions.
RESUMEN
Purpose: Arterial stiffness and steno-occlusion of the lower-extremity can result from many vascular lesions, including acute thromboembolisms, soft plaques, calcified plaques, or inflammatory disease. Ultrasound (US) elastography measures the tissue deformation response to compression and displays tissue stiffness. This study aimed to evaluate the characteristics of arterial lesions in the lower extremities using US elastography. Materials and Methods: We retrospectively analyzed the data of 20 patients who visited our institute for arterial disease treatment between May 2016 and November 2017. An US examination with B-mode and strain elastography (SE) was performed of four different lesion types at 45 sites: acute and subacute thromboembolisms, soft plaques, calcified plaques, and thromboangiitis obliterans lesions (TAOs). During SE, stress was externally applied by the operator using the transducer. Strain ratio (SR) was calculated as the fraction of the average strain in the reference area divided by the average strain in the lesion. The SR was compared among different lesion types, with the accompanying vein as the reference region of interest. Results: The strain was highest in the soft plaques (0.63%±0.23%), followed by the TAOs (0.45%±0.11%), calcified plaques (0.44%±0.13%), and acute thromboembolisms (0.34%±0.23%), which were statistically significant (P=0.026). However, the mean SR was highest for the calcified plaques (2.33%±0.80%), followed by the TAOs (1.63%±0.40%), acute thromboembolisms (1.60%±0.48%), and soft plaques (1.51±0.39), and which were statistically significant (P=0.013). Conclusion: Despite several limitations, vascular elastography may be useful for differentiating between lesion types in peripheral arterial disease.
RESUMEN
BACKGROUND AND OBJECTIVE: The protocol for placing anatomical side markers (L/R markers) in chest radiographs varies from one hospital or department to another. However, the markers have strong signals that can be useful for deep learning-based classifier to predict diseases. We aimed to enhance the performance of a deep learning-based classifiers in multi-center datasets by inpainting the L/R markers. METHODS: The L/R marker was detected with using the EfficientDet detection network; only the detected regions were inpainted using a generative adversarial network (GAN). To analyze the effect of the inpainting in detail, deep learning-based classifiers were trained using original images, marker-inpainted images, and original images clipped using the min-max value of the marker-inpainted images. Binary classification, multi-class classification, and multi-task learning with segmentation and classification were developed and evaluated. Furthermore, the performances of the network on internal and external validation datasets were compared using DeLong's test for two correlated receiver operating characteristic (ROC) curves in binary classification and Stuart-Maxwell test for marginal homogeneity in multi-class classification and multi-task learning. In addition, the qualitative results of activation maps were evaluated using the gradient-class activation map (Grad-CAM). RESULTS: Marker-inpainting preprocessing improved the classification performances. In the binary classification based on the internal validation, the area under the curves (AUCs) and accuracies were 0.950 and 0.900 for the model trained on the min-max clipped images and 0.911 and 0.850 for the model trained on the original images, respectively (P-value=0.006). In the external validation, the AUCs and accuracies were 0.858 and 0.677 for the model trained using the inpainted images and 0.723 and 0.568 for the model trained using the original images (P-value<0.001), respectively. In addition, the models trained using the marker inpainted images showed the best performance in multi-class classification and multi-task learning. Furthermore, the activation maps obtained using the Grad-CAM improved with the proposed method. The 5-fold validation results also showed improvement trend according to the preprocessing strategies. CONCLUSIONS: Inpainting an L/R marker significantly enhanced the classifier's performance and robustness, especially in internal and external studies, which could be useful in developing a more robust and accurate deep learning-based classifier for multi-center trials. The code for detection is available at: https://github.com/mi2rl/MI2RLNet. And the code for inpainting is available at: https://github.com/mi2rl/L-R-marker-inpainting.
Asunto(s)
Aprendizaje Profundo , Área Bajo la Curva , Curva ROC , RadiografíaRESUMEN
With the recent development of deep learning, the classification and segmentation tasks of computer-aided diagnosis (CAD) using non-contrast head computed tomography (NCCT) for intracranial hemorrhage (ICH) has become popular in emergency medical care. However, a few challenges remain, such as the difficulty of training due to the heterogeneity of ICH, the requirement for high performance in both sensitivity and specificity, patient-level predictions demanding excessive costs, and vulnerability to real-world external data. In this study, we proposed a supervised multi-task aiding representation transfer learning network (SMART-Net) for ICH to overcome these challenges. The proposed framework consists of upstream and downstream components. In the upstream, a weight-shared encoder of the model is trained as a robust feature extractor that captures global features by performing slice-level multi-pretext tasks (classification, segmentation, and reconstruction). Adding a consistency loss to regularize discrepancies between classification and segmentation heads has significantly improved representation and transferability. In the downstream, the transfer learning was conducted with a pre-trained encoder and 3D operator (classifier or segmenter) for volume-level tasks. Excessive ablation studies were conducted and the SMART-Net was developed with optimal multi-pretext task combinations and a 3D operator. Experimental results based on four test sets (one internal and two external test sets that reflect a natural incidence of ICH, and one public test set with a relatively small amount of ICH cases) indicate that SMART-Net has better robustness and performance in terms of volume-level ICH classification and segmentation over previous methods. All code is available at https://github.com/babbu3682/SMART-Net.
Asunto(s)
Hemorragias Intracraneales , Tomografía Computarizada por Rayos X , Diagnóstico por Computador , Humanos , Procesamiento de Imagen Asistido por Computador/métodos , Hemorragias Intracraneales/diagnóstico por imagen , Sensibilidad y EspecificidadRESUMEN
STUDY OBJECTIVES: To evaluate whether obstructive sleep apnea (OSA) and its severity are related to dyslipidemia and alanine transaminase elevation as a marker of nonalcoholic fatty liver disease in children. METHODS: The data collected from polysomnography, laboratory measurements (lipid profile and liver enzyme), and body mass index in children aged 0-18 years who visited the pediatric department between 2012 and 2018 were retrospectively analyzed. RESULTS: There were a total of 273 participants in the study (ages 0-6 years, 7-12 years, and 13-18 years: 61.9%, 26.4%, and 11.7%, respectively). In the ages 7-12 and 13-18 years groups, obesity was strongly associated with OSA severity (Cramer's V = 0.498, P < .001). High-density lipoprotein cholesterol levels were significantly lower in the OSA group than in the non-OSA group, irrespective of the presence of obesity. In addition, high-density lipoprotein cholesterol levels were significantly different between the OSA severity groups after adjusting for body mass index (P = .000). In participants who were obese, moderate and severe OSA were associated with alanine transaminase elevation (P = .023 and P = .045, respectively). CONCLUSIONS: This study suggests that OSA may be an independent risk factor for dyslipidemia and that OSA and obesity have a synergistic effect on alanine transaminase elevation. Early diagnosis and treatment of OSA from childhood, especially in obese children, will reduce metabolic complications. CITATION: Kang EK, Jang MJ, Kim KD, Ahn YM. The association of obstructive sleep apnea with dyslipidemia in Korean children and adolescents: a single-center, cross-sectional study. J Clin Sleep Med. 2021;17(8):1599-1605.
Asunto(s)
Dislipidemias , Obesidad Infantil , Apnea Obstructiva del Sueño , Adolescente , Niño , Preescolar , Estudios Transversales , Dislipidemias/complicaciones , Dislipidemias/epidemiología , Humanos , Lactante , Recién Nacido , República de Corea/epidemiología , Estudios Retrospectivos , Apnea Obstructiva del Sueño/complicaciones , Apnea Obstructiva del Sueño/epidemiologíaRESUMEN
BACKGROUND: Factors determining bisphosphonate compliance are not fully understood. We examined fluctuations in oral bisphosphonate dosing intervals to gauge therapeutic compliance in patients with osteoporosis. MATERIALS AND METHODS: Hospital data accruing between 2010 and 2017 were accessed to retrospectively study patients ≥50 years old (N=1873), each prescribed bisphosphonate at initial diagnosis of osteoporosis. The medication possession ratio (MPR), calculated as total days supplied divided by length of follow-up, served to measure therapeutic compliance. We compared MPRs of various prescription patterns (daily, weekly, monthly, and switch [ie, ≥1 change in pattern] groups). We also analyzed the impact of age, sex, fracture history, surgical history, and comorbidities. Multiple regression analysis was ultimately performed, using MPR as a dependent variable. RESULTS: In our cohort (mean follow-up=5.7±2.4 years), once weekly dosing was the most common prescription pattern (1223/1873, 65.3%), as opposed to monthly (366/1873, 19.5%) or daily (164/1873, 8.8%) dosing. A total of 120 patients (6.4%) comprising the switch group changed dosing patterns during the study period. MPR was significantly higher in the switch group (32.8±22.7) than in the other three groups (daily, 21.9±25.9; weekly, 22.7±27.3; monthly, 23.2±27.7). In multiple regression analysis, younger age (P<0.001), female sex (P=0.004), and switching of prescription pattern (decrease or increase frequency) were factors significantly associated with higher MPR, signaling better compliance. CONCLUSION: Better bisphosphonate compliance was associated with physician-modified dosing patterns. We therefore recommend adjustments of prescription intervals in poorly compliant patients requiring long-term treatment.
RESUMEN
This paper presents three methods of input voltage signals that allow low voltage operation of an electrovibration display while preserving the perceptual feel and strength of electrovibration stimuli. The first method uses the amplitude modulation of a high-frequency carrier-signal. The second method uses a dc-offset, and the third method uses a combination of the two methods. The performance of the three methods was evaluated by a physical experiment that measured and analyzed static (dc-component) and dynamic (vibratory component) friction forces and two subsequent psychophysical studies. The physical experiment showed that only the dc -offset method enabled a statistically significant increase in the static friction force between the fingertip and the surface of the electrovibration display. The static friction increase was closely related to the root mean square of input voltage level. In contrast, all of the three methods increased the dynamic friction force significantly, which was deemed to be related to the high frequency effect validated in the previous literature. The first psychophysical study showed that the three proposed methods can significantly reduce the peak-to-peak (p-p) amplitude of an input voltage signal while generating perceptually equally strong electrovibrations to that produced by the conventional method. Using lower p-p voltage has the merits of a simpler electrical circuit and less electromagnetic noise, saving the overall system cost. Further, the perceived intensity of electrovibration was more correlated to the dynamic friction force than the static friction force. The second psychophysical study was a discrimination experiment, and it demonstrated that all the three proposed methods and the conventional method can provide perceptually similar stimuli despite their different signal forms and voltage amplitudes. Our experimental investigation allowed us to conclude that the dc-offset method is the best way to lower the driving voltage of an electrovibration display while providing perceptually equivalent electrovibrations.
Asunto(s)
Discriminación en Psicología/fisiología , Fenómenos Electromagnéticos , Dedos , Fricción , Psicofísica , Percepción del Tacto/fisiología , Vibración , Adulto , Equipos y Suministros Eléctricos , Femenino , Humanos , Masculino , Adulto JovenRESUMEN
Electrovibration is a type of surface haptics that can modulate lateral forces acting between a fingertip and a touch surface. Electrovibration is fast, consumes little power, and does not involve the use of any mechanical actuators. However, it suffers from problems such as nonuniform perceived intensity due to varying environmental impedances, as well as possible electric shock, which have to be solved for commercialization. In this paper, a current feedback method is proposed to provide uniform intensity of electrovibration, regardless of the varying environmental impedances. The proposed method can also prevent electric shock. To show the effectiveness of the proposed method, a hardware prototype was developed and a user study was conducted. The user study result shows that the proposed current control method can provide significantly more uniform perceived intensity of electrovibration as compared with the conventional voltage control method.