Application of ultrasound-based radiomics technology in fetal-lung-texture analysis in pregnancies complicated by gestational diabetes and/or pre-eclampsia

Deep learning approaches for automated classification of neonatal lung ultrasound with assessment of human-to-AI interrater agreement

Neonatal respiratory disorders pose significant challenges in clinical settings, often requiring rapid and accurate diagnostic solutions for effective management. Lung ultrasound (LUS) has emerged as a promising tool to evaluate respiratory conditions in neonates. This evaluation is mainly based on the interpretation of visual patterns (horizontal artifacts, vertical artifacts, and consolidations). Automated interpretation of these patterns can assist clinicians in their evaluations. However, developing AI-based solutions for this purpose is challenging, primarily due to the lack of annotated data and inherent subjectivity in expert interpretations. This study aims to propose an automated solution for the reliable interpretation of patterns in LUS videos of newborns. We employed two distinct strategies. The first strategy is a frame-to-video-level approach that computes frame-level predictions from deep learning (DL) models trained from scratch (F2V-TS) along with fine-tuning pre-trained models (F2V-FT) followed by aggregation of those predictions for video-level evaluation. The second strategy is a direct video classification approach (DV) for evaluating LUS data. To evaluate our methods, we used LUS data from 34 neonatal patients comprising of 70 exams with annotations provided by three expert human operators (3HOs). Results show that within the frame-to-video-level approach, F2V-FT achieved the best performance with an accuracy of 77% showing moderate agreement with the 3HOs. while the direct video classification approach resulted in an accuracy of 72%, showing substantial agreement with the 3HOs, our proposed study lays down the foundation for reliable AI-based solutions for newborn LUS data evaluation.

Evaluation of placenta and fetal lung using shear wave elastography in gestational diabetes mellitus: An innovative approach

PURPOSE

This study aims to investigate placental and fetal lung stiffness in pregnant women with and without gestational diabetes, considering the well-established delay in fetal lung maturation associated with gestational diabetes.

MATERIALS AND METHODS

This prospective cohort study was conducted at a tertiary center and included pregnant women who underwent a 75-gram oral glucose tolerance test between 24-28 weeks of gestation. Elastography measurements were performed using point shear wave elastography (pSWE).

RESULTS

The study included 60 pregnant women diagnosed with gestational diabetes and 60 pregnant women in the control group. The SWE velocity of the peripheral placenta, central placenta, and lung was higher in the gestational diabetes group compared to the control group. Furthermore, the SWE velocity of the peripheral placenta, central placenta, and lung was higher in newborns with neonatal respiratory morbidity. Based on the ROC analysis of patients with gestational diabetes, the AUC for lung SWE velocity was 0.88 (cut-off 12.4 kPa, 95% CI: 0.77-0.99, p<0.001) with a sensitivity of 71.4% and specificity of 95.6% for predicting neonatal respiratory morbidity.

CONCLUSION

Fetal placental and lung stiffness increase in fetuses of pregnant women with diabetes. Moreover, higher fetal lung stiffness during the fetal period is associated with increased neonatal respiratory morbidity.

Ultrasound-Based Radiomics for the Classification of Henoch-Schönlein Purpura Nephritis in Children

Henoch-Schönlein purpura nephritis (HSPN) is one of the most common kidney diseases in children. The current diagnosis and classification of HSPN depend on pathological biopsy, which is seriously limited by its invasive and high-risk nature. The aim of the study was to explore the potential of radiomics model for evaluating the histopathological classification of HSPN based on the ultrasound (US) images. A total of 440 patients with Henoch-Schönlein purpura nephritis proved by biopsy were analyzed retrospectively. They were grouped according to two histopathological categories: those without glomerular crescent formation (ISKDC grades I-II) and those with glomerular crescent formation (ISKDC grades III-V). The patients were randomly assigned to either a training cohort (n = 308) or a validation cohort (n = 132) with a ratio of 7:3. The sonologist manually drew the regions of interest (ROI) on the ultrasound images of the right kidney including the cortex and medulla. Then, the ultrasound radiomics features were extracted using the Pyradiomics package. The dimensions of radiomics features were reduced by Spearman correlation coefficients and least absolute shrinkage and selection operator (LASSO) method. Finally, three radiomics models using k-nearest neighbor (KNN), logistic regression (LR), and support vector machine (SVM) were established, respectively. The predictive performance of such classifiers was assessed with receiver operating characteristic (ROC) curve. 105 radiomics features were extracted from derived US images of each patient and 14 features were ultimately selected for the machine learning analysis. Three machine learning models including k-nearest neighbor (KNN), logistic regression (LR), and support vector machine (SVM) were established for HSPN classification. Of the three classifiers, the SVM classifier performed the best in the validation cohort [area under the curve (AUC) =0.870 (95% CI, 0.795-0.944), sensitivity = 0.706, specificity = 0.950]. The US-based radiomics had good predictive value for HSPN classification, which can be served as a noninvasive tool to evaluate the severity of renal pathology and crescentic formation in children with HSPN.

Classification of lung pathologies in neonates using dual-tree complex wavelet transform

INTRODUCTION

Undiagnosed and untreated lung pathologies are among the leading causes of neonatal deaths in developing countries. Lung Ultrasound (LUS) has been widely accepted as a diagnostic tool for neonatal lung pathologies due to its affordability, portability, and safety. However, healthcare institutions in developing countries lack well-trained clinicians to interpret LUS images, which limits the use of LUS, especially in remote areas. An automated point-of-care tool that could screen and capture LUS morphologies associated with neonatal lung pathologies could aid in rapid and accurate diagnosis.

METHODS

We propose a framework for classifying the six most common neonatal lung pathologies using spatially localized line and texture patterns extracted via 2D dual-tree complex wavelet transform (DTCWT). We acquired 1550 LUS images from 42 neonates with varying numbers of lung pathologies. Furthermore, we balanced our data set to avoid bias towards a pathology class.

RESULTS

Using DTCWT and clinical features as inputs to a linear discriminant analysis (LDA), our approach achieved a per-image cross-validated classification accuracy of 74.39% for the imbalanced data set. Our classification accuracy improved to 92.78% after balancing our data set. Moreover, our proposed framework achieved a maximum per-subject cross-validated classification accuracy of 64.97% with an imbalanced data set while using a balanced data set improves its classification accuracy up to 81.53%.

CONCLUSION

Our work could aid in automating the diagnosis of lung pathologies among neonates using LUS. Rapid and accurate diagnosis of lung pathologies could help to decrease neonatal deaths in healthcare institutions that lack well-trained clinicians, especially in developing countries.

Enhancing Fetal Anomaly Detection in Ultrasonography Images: A Review of Machine Learning-Based Approaches

Fetal development is a critical phase in prenatal care, demanding the timely identification of anomalies in ultrasound images to safeguard the well-being of both the unborn child and the mother. Medical imaging has played a pivotal role in detecting fetal abnormalities and malformations. However, despite significant advances in ultrasound technology, the accurate identification of irregularities in prenatal images continues to pose considerable challenges, often necessitating substantial time and expertise from medical professionals. In this review, we go through recent developments in machine learning (ML) methods applied to fetal ultrasound images. Specifically, we focus on a range of ML algorithms employed in the context of fetal ultrasound, encompassing tasks such as image classification, object recognition, and segmentation. We highlight how these innovative approaches can enhance ultrasound-based fetal anomaly detection and provide insights for future research and clinical implementations. Furthermore, we emphasize the need for further research in this domain where future investigations can contribute to more effective ultrasound-based fetal anomaly detection.

Evolving the Era of 5D Ultrasound? A Systematic Literature Review on the Applications for Artificial Intelligence Ultrasound Imaging in Obstetrics and Gynecology

Artificial intelligence (AI) has gained prominence in medical imaging, particularly in obstetrics and gynecology (OB/GYN), where ultrasound (US) is the preferred method. It is considered cost effective and easily accessible but is time consuming and hindered by the need for specialized training. To overcome these limitations, AI models have been proposed for automated plane acquisition, anatomical measurements, and pathology detection. This study aims to overview recent literature on AI applications in OB/GYN US imaging, highlighting their benefits and limitations. For the methodology, a systematic literature search was performed in the PubMed and Cochrane Library databases. Matching abstracts were screened based on the PICOS (Participants, Intervention or Exposure, Comparison, Outcome, Study type) scheme. Articles with full text copies were distributed to the sections of OB/GYN and their research topics. As a result, this review includes 189 articles published from 1994 to 2023. Among these, 148 focus on obstetrics and 41 on gynecology. AI-assisted US applications span fetal biometry, echocardiography, or neurosonography, as well as the identification of adnexal and breast masses, and assessment of the endometrium and pelvic floor. To conclude, the applications for AI-assisted US in OB/GYN are abundant, especially in the subspecialty of obstetrics. However, while most studies focus on common application fields such as fetal biometry, this review outlines emerging and still experimental fields to promote further research.

Artificial intelligence in obstetric ultrasound: A scoping review

The objective is to summarize the current use of artificial intelligence (AI) in obstetric ultrasound. PubMed, Cochrane Library, and ClinicalTrials.gov databases were searched using the following keywords "neural networks", OR "artificial intelligence", OR "machine learning", OR "deep learning", AND "obstetrics", OR "obstetrical", OR "fetus", OR "foetus", OR "fetal", OR "foetal", OR "pregnancy", or "pregnant", AND "ultrasound" from inception through May 2022. The search was limited to the English language. Studies were eligible for inclusion if they described the use of AI in obstetric ultrasound. Obstetric ultrasound was defined as the process of obtaining ultrasound images of a fetus, amniotic fluid, or placenta. AI was defined as the use of neural networks, machine learning, or deep learning methods. The authors' search identified a total of 127 papers that fulfilled our inclusion criteria. The current uses of AI in obstetric ultrasound include first trimester pregnancy ultrasound, assessment of placenta, fetal biometry, fetal echocardiography, fetal neurosonography, assessment of fetal anatomy, and other uses including assessment of fetal lung maturity and screening for risk of adverse pregnancy outcomes. AI holds the potential to improve the ultrasound efficiency, pregnancy outcomes in low resource settings, detection of congenital malformations and prediction of adverse pregnancy outcomes.

Reproducibility of 2D versus 3D radiomics for quantitative assessment of fetal lung development: a retrospective fetal MRI study

PURPOSE

To investigate the reproducibility of radiomics features extracted from two-dimensional regions of interest (2D ROIs) versus whole lung (3D) ROIs in repeated in-vivo fetal magnetic resonance imaging (MRI) acquisitions.

METHODS

Thirty fetal MRI scans including two axial T2-weighted acquisitions of the lungs were analysed. 2D (lung at the level of the carina) and 3D (whole lung) ROIs were manually segmented using ITK-Snap. Ninety-five radiomics features were extracted from 2 and 3D ROIs in initial and repeat acquisitions using Pyradiomics. Radiomics feature intra-class correlation coefficients (ICC) were calculated between 2 and 3D ROIs in the initial acquisition, and between 2 and 3D ROIs in repeated acquisitions, respectively.

RESULTS

MRI data of 11 (36.7%) female and 19 (63.3%) male fetuses acquired at a median 25 + 0 gestational weeks plus days (GW) (interquartile range [IQR] 23 + 4 - 27 + 0 GW) were assessed. Median radiomics feature ICC between 2 and 3D ROIs in the initial MRI acquisition was 0.733 (IQR 0.313-0.814, range 0.018-0.970). ICCs between radiomics features extracted using 3D ROIs in initial and repeat acquisitions (median 0.908 [IQR 0.824-0.929, range 0.335-0.996]) were significantly higher compared to 2D ROIs (0.771 [0.699-0.835, 0.048-0.965]) (p < 0.001).

CONCLUSION

Fetal MRI radiomics features extracted from 3D whole lung segmentation masks showed significantly higher reproducibility across repeat acquisitions compared to 2D ROIs. Therefore, fetal MRI whole lung radiomics features are robust diagnostic and potentially prognostic tools in the image-based in-vivo quantitative assessment of lung development.

Musculoskeletal Ultrasound Image-Based Radiomics for the Diagnosis of Achilles Tendinopathy in Skiers

OBJECTIVES

Our study aimed to develop and validate an efficient ultrasound image-based radiomic model for determining the Achilles tendinopathy in skiers.

METHODS

A total of 88 feet of skiers clinically diagnosed with unilateral chronic Achilles tendinopathy and 51 healthy feet were included in our study. According to the time order of enrollment, the data were divided into a training set (n = 89) and a test set (n = 50). The regions of interest (ROIs) were segmented manually, and 833 radiomic features were extracted from red, green, blue color channels and grayscale of ROIs using Pyradiomics, respectively. Three feature selection and three machine learning modeling algorithms were implemented respectively, for determining the optimal radiomics pipeline. Finally, the area under the receiver operating characteristic curve (AUC), consistency analysis, and decision analysis were used to evaluate the diagnostic performance.

RESULTS

By comparing nine radiomics analysis strategies of three color channels and grayscale, the radiomic model under the green channel obtained the best diagnostic performance, using the Random Forest selection and Support Vector Machine modeling, which was selected as the final machine learning model. All the selected radiomic features were significantly associated with the Achilles tendinopathy (P < .05). The radiomic model had a training AUC of 0.98, a test AUC of 0.99, a sensitivity of 0.90, and a specificity of 1, which could bring sufficient clinical net benefits.

CONCLUSIONS

Ultrasound image-based radiomics achieved high diagnostic performance, which could be used as an intelligent auxiliary tool for the diagnosis of Achilles tendinopathy.

Fetal MRI radiomics: non-invasive and reproducible quantification of human lung maturity

OBJECTIVES

To assess the reproducibility of radiomics features extracted from the developing lung in repeated in-vivo fetal MRI acquisitions.

METHODS

In-vivo MRI (1.5 Tesla) scans of 30 fetuses, each including two axial and one coronal T2-weighted sequences of the whole lung with all other acquisition parameters kept constant, were retrospectively identified. Manual segmentation of the lungs was performed using ITK-Snap. One hundred radiomics features were extracted from fetal lung MRI data using Pyradiomics, resulting in 90 datasets. Intra-class correlation coefficients (ICC) of radiomics features were calculated between baseline and repeat axial acquisitions and between baseline axial and coronal acquisitions.

RESULTS

MRI data of 30 fetuses (12 [40%] females, 18 [60%] males) at a median gestational age of 24 + 5 gestational weeks plus days (GW) (interquartile range [IQR] 3 + 3 GW, range 21 + 1 to 32 + 6 GW) were included. Median ICC of radiomics features between baseline and repeat axial MR acquisitions was 0.92 (IQR 0.13, range 0.33 to 1), with 60 features exhibiting excellent (ICC > 0.9), 27 good (> 0.75-0.9), twelve moderate (0.5-0.75), and one poor (ICC < 0.5) reproducibility. Median ICC of radiomics features between baseline axial and coronal MR acquisitions was 0.79 (IQR 0.15, range 0.2 to 1), with 20 features exhibiting excellent, 47 good, 29 moderate, and four poor reproducibility.

CONCLUSION

Standardized in-vivo fetal MRI allows reproducible extraction of lung radiomics features. In the future, radiomics analysis may improve diagnostic and prognostic yield of fetal MRI in normal and pathologic lung development.

KEY POINTS

• Non-invasive fetal MRI acquired using a standardized protocol allows reproducible extraction of radiomics features from the developing lung for objective tissue characterization. • Alteration of imaging plane between fetal MRI acquisitions has a negative impact on lung radiomics feature reproducibility. • Fetal MRI radiomics features reflecting the microstructure and shape of the fetal lung could complement observed-to-expected lung volume in the prediction of postnatal outcome and optimal treatment of fetuses with abnormal lung development in the future.

Ultrasound-based radiomics for the evaluation of fetal rat lung maturity a non-invasive assessment method (Ultrasound-based radiomics in fetal rat lung)

OBJECTIVE

To establish a classification model for the evaluation of rat fetal lung maturity (FLM) using radiomics technology.

METHOD

A total of 430 high-throughput features were extracted per fetal lung image from 134 fetal lung ultrasound images (four-cardiac-chamber views) of 67 Sprague-Dawley (SD) fetal rats with gestational age (GA) of 16-21 days. The detection of fetal lung tissues included histopathological staining and the expression of the surface protein (SP) SP-A, SP-B, and SP-C. A machine learning classification model was established by a support vector machine based on histopathological results to analyze the relationship between fetal lung texture characteristics and FLM.

RESULTS

The rat fetal lungs were divided into two groups: terminal sac period (SD1) and canalicular period (SD2). The mRNA transcription and protein expression level of SP-C protein were significantly higher in the SD1 group than in the SD2 group (P < 0.05). The diagnostic performance of the rat FLM classification model was measured as follows: area under the receiver operating characteristic curve (AUC), 0.93 (training set) and 0.89 (validation set); sensitivity, 89.26% (training set) and 87.10% (validation set); specificity, 85.87% (training set) and 79.17% (validation set); accuracy, 87.79% (training set) and 83.64% (validation set).

CONCLUSION

Ultrasound-based radiomics technology can be used to evaluate the FLM of rats, which lays a foundation for further research on this technology in human fetal lungs. This article is protected by copyright. All rights reserved.

Ultrasound-based radiomics technology in fetal lung texture analysis prediction of neonatal respiratory morbidity

To develop a novel method for predicting neonatal respiratory morbidity (NRM) by ultrasound-based radiomics technology. In this retrospective study, 430 high-throughput features per fetal-lung image were extracted from 295 fetal lung ultrasound images (four-chamber view) in 295 single pregnancies. Images had been obtained between 28⁺³ and 37⁺⁶ weeks of gestation within 72 h before delivery. A machine-learning model built by RUSBoost (Random under-sampling with AdaBoost) architecture was created using 20 radiomics features extracted from the images and 2 clinical features (gestational age and pregnancy complications) to predict the possibility of NRM. Of the 295 standard fetal lung ultrasound images included, 210 in the training set and 85 in the testing set. The overall performance of the neonatal respiratory morbidity prediction model achieved AUC of 0.88 (95% CI 0.83–0.92) in the training set and 0.83 (95% CI 0.79–0.97) in the testing set, sensitivity of 84.31% (95% CI 79.06–89.44%) in the training set and 77.78% (95% CI 68.30–87.43%) in the testing set, specificity of 81.13% (95% CI 78.16–84.07%) in the training set and 82.09% (95% CI 77.65–86.62%) in the testing set, and accuracy of 81.90% (95% CI 79.34–84.41%) in the training set and 81.18% (95% CI 77.33–85.12%) in the testing set. Ultrasound-based radiomics technology can be used to predict NRM. The results of this study may provide a novel method for non-invasive approaches for the prenatal prediction of NRM.

Use of Machine Learning and Routine Laboratory Tests for Diabetes Mellitus Screening

Most patients with diabetes mellitus are asymptomatic, which leads to delayed and more complex treatment. At the same time, most individuals are routinely subjected to standard clinical laboratory examinations, which create large health datasets over a lifetime. Computer processing has been used to search for health anomalies and predict diseases using clinical examinations. This work studied machine learning models to support the screening of diabetes through routine laboratory tests using data from laboratory tests of 62,496 patients. The classification and regression models used were the K-nearest neighbor, support vector machines, Bayes naïve, random forest models, and artificial neural networks. Glycated hemoglobin, a test used for diabetes diagnosis, was used as the target. Regression models calculated glycated hemoglobin directly and were later classified. The performance of classification computer models has been studied under various subdataset partitions and combinations (e.g., healthy, prediabetic, and diabetes, as well as no healthy and no diabetes). The best single performance was achieved with the artificial neural network model when detecting prediabetes or diabetes. The artificial neural network classification model scored 78.1%, 78.7%, and 78.4% for sensitivity, precision, and F1 scores, respectively, when identifying no healthy group. Other models also had good results, depending on what is desired. Machine learning-based models can predict glycated hemoglobin values from routine laboratory tests and can be used as a screening tool to refer a patient for further testing.

Deep Learning With Radiomics for Disease Diagnosis and Treatment: Challenges and Potential

The high-throughput extraction of quantitative imaging features from medical images for the purpose of radiomic analysis, i.e., radiomics in a broad sense, is a rapidly developing and emerging research field that has been attracting increasing interest, particularly in multimodality and multi-omics studies. In this context, the quantitative analysis of multidimensional data plays an essential role in assessing the spatio-temporal characteristics of different tissues and organs and their microenvironment. Herein, recent developments in this method, including manually defined features, data acquisition and preprocessing, lesion segmentation, feature extraction, feature selection and dimension reduction, statistical analysis, and model construction, are reviewed. In addition, deep learning-based techniques for automatic segmentation and radiomic analysis are being analyzed to address limitations such as rigorous workflow, manual/semi-automatic lesion annotation, and inadequate feature criteria, and multicenter validation. Furthermore, a summary of the current state-of-the-art applications of this technology in disease diagnosis, treatment response, and prognosis prediction from the perspective of radiology images, multimodality images, histopathology images, and three-dimensional dose distribution data, particularly in oncology, is presented. The potential and value of radiomics in diagnostic and therapeutic strategies are also further analyzed, and for the first time, the advances and challenges associated with dosiomics in radiotherapy are summarized, highlighting the latest progress in radiomics. Finally, a robust framework for radiomic analysis is presented and challenges and recommendations for future development are discussed, including but not limited to the factors that affect model stability (medical big data and multitype data and expert knowledge in medical), limitations of data-driven processes (reproducibility and interpretability of studies, different treatment alternatives for various institutions, and prospective researches and clinical trials), and thoughts on future directions (the capability to achieve clinical applications and open platform for radiomics analysis).

Novel artificial intelligence approach for automatic differentiation of fetal occiput anterior and non-occiput anterior positions during labor

OBJECTIVES

To describe a newly developed machine-learning (ML) algorithm for the automatic recognition of fetal head position using transperineal ultrasound (TPU) during the second stage of labor and to describe its performance in differentiating between occiput anterior (OA) and non-OA positions.

METHODS

This was a prospective cohort study including singleton term (> 37 weeks of gestation) pregnancies in the second stage of labor, with a non-anomalous fetus in cephalic presentation. Transabdominal ultrasound was performed to determine whether the fetal head position was OA or non-OA. For each case, one sonographic image of the fetal head was then acquired in an axial plane using TPU and saved for later offline analysis. Using the transabdominal sonographic diagnosis as the gold standard, a ML algorithm based on a pattern-recognition feed-forward neural network was trained on the TPU images to discriminate between OA and non-OA positions. In the training phase, the model tuned its parameters to approximate the training data (i.e. the training dataset) such that it would identify correctly the fetal head position, by exploiting geometric, morphological and intensity-based features of the images. In the testing phase, the algorithm was blinded to the occiput position as determined by transabdominal ultrasound. Using the test dataset, the ability of the ML algorithm to differentiate OA from non-OA fetal positions was assessed in terms of diagnostic accuracy. The F₁ -score and precision-recall area under the curve (PR-AUC) were calculated to assess the algorithm's performance. Cohen's kappa (κ) was calculated to evaluate the agreement between the algorithm and the gold standard.

RESULTS

Over a period of 24 months (February 2018 to January 2020), at 15 maternity hospitals affiliated to the International Study group on Labor ANd Delivery Sonography (ISLANDS), we enrolled into the study 1219 women in the second stage of labor. On the basis of transabdominal ultrasound, they were classified as OA (n = 801 (65.7%)) or non-OA (n = 418 (34.3%)). From the entire cohort (OA and non-OA), approximately 70% (n = 824) of the patients were assigned randomly to the training dataset and the rest (n = 395) were used as the test dataset. The ML-based algorithm correctly classified the fetal occiput position in 90.4% (357/395) of the test dataset, including 224/246 with OA (91.1%) and 133/149 with non-OA (89.3%) fetal head position. Evaluation of the algorithm's performance gave an F₁ -score of 88.7% and a PR-AUC of 85.4%. The algorithm showed a balanced performance in the recognition of both OA and non-OA positions. The robustness of the algorithm was confirmed by high agreement with the gold standard (κ = 0.81; P < 0.0001).

CONCLUSIONS

This newly developed ML-based algorithm for the automatic assessment of fetal head position using TPU can differentiate accurately, in most cases, between OA and non-OA positions in the second stage of labor. This algorithm has the potential to support not only obstetricians but also midwives and accoucheurs in the clinical use of TPU to determine fetal occiput position in the labor ward. © 2021 International Society of Ultrasound in Obstetrics and Gynecology.

Establish a normal fetal lung gestational age grading model and explore the potential value of deep learning algorithms in fetal lung maturity evaluation

Background:

Prenatal evaluation of fetal lung maturity (FLM) is a challenge, and an effective non-invasive method for prenatal assessment of FLM is needed. The study aimed to establish a normal fetal lung gestational age (GA) grading model based on deep learning (DL) algorithms, validate the effectiveness of the model, and explore the potential value of DL algorithms in assessing FLM.

Methods:

A total of 7013 ultrasound images obtained from 1023 normal pregnancies between 20 and 41 + 6 weeks were analyzed in this study. There were no pregnancy-related complications that affected fetal lung development, and all infants were born without neonatal respiratory diseases. The images were divided into three classes based on the gestational week: class I: 20 to 29 + 6 weeks, class II: 30 to 36 + 6 weeks, and class III: 37 to 41 + 6 weeks. There were 3323, 2142, and 1548 images in each class, respectively. First, we performed a pre-processing algorithm to remove irrelevant information from each image. Then, a convolutional neural network was designed to identify different categories of fetal lung ultrasound images. Finally, we used ten-fold cross-validation to validate the performance of our model. This new machine learning algorithm automatically extracted and classified lung ultrasound image information related to GA. This was used to establish a grading model. The performance of the grading model was assessed using accuracy, sensitivity, specificity, and receiver operating characteristic curves.

Results:

A normal fetal lung GA grading model was established and validated. The sensitivity of each class in the independent test set was 91.7%, 69.8%, and 86.4%, respectively. The specificity of each class in the independent test set was 76.8%, 90.0%, and 83.1%, respectively. The total accuracy was 83.8%. The area under the curve (AUC) of each class was 0.982, 0.907, and 0.960, respectively. The micro-average AUC was 0.957, and the macro-average AUC was 0.949.

Conclusions:

The normal fetal lung GA grading model could accurately identify ultrasound images of the fetal lung at different GAs, which can be used to identify cases of abnormal lung development due to gestational diseases and evaluate lung maturity after antenatal corticosteroid therapy. The results indicate that DL algorithms can be used as a non-invasive method to predict FLM.

Germline BRCA 1-2 status prediction through ovarian ultrasound images radiogenomics: a hypothesis generating study (PROBE study)

Radiogenomics is a specific application of radiomics where imaging features are linked to genomic profiles. We aim to develop a radiogenomics model based on ovarian US images for predicting germline BRCA1/2 gene status in women with healthy ovaries. From January 2013 to December 2017 a total of 255 patients addressed to germline BRCA1/2 testing and pelvic US documenting normal ovaries, were retrospectively included. Feature selection for univariate analysis was carried out via correlation analysis. Multivariable analysis for classification of germline BRCA1/2 status was then carried out via logistic regression, support vector machine, ensemble of decision trees and automated machine learning pipelines. Data were split into a training (75%) and a testing (25%) set. The four strategies obtained a similar performance in terms of accuracy on the testing set (from 0.54 of logistic regression to 0.64 of the auto-machine learning pipeline). Data coming from one of the tested US machine showed generally higher performances, particularly with the auto-machine learning pipeline (testing set specificity 0.87, negative predictive value 0.73, accuracy value 0.72 and 0.79 on training set). The study shows that a radiogenomics model on machine learning techniques is feasible and potentially useful for predicting gBRCA1/2 status in women with healthy ovaries.