Quantitative Ultrasound Texture Analysis for Differentiating Preterm From Term Fetal Lungs

Machine Learning on Ultrasound Texture Analysis Data for Characterizing of Salivary Glandular Tumors: A Feasibility Study

BACKGROUND

Objective quantitative texture characteristics may be helpful in salivary glandular tumor differential diagnosis. This study uses machine learning (ML) to explore and validate the performance of ultrasound (US) texture features in diagnosing salivary glandular tumors.

MATERIAL AND METHODS

122 patients with salivary glandular tumors, including 71 benign and 51 malignant tumors, are enrolled. Representative brightness mode US pictures are selected for further Gray Level Co-occurrence Matrix (GLCM) texture analysis. We use a t-test to test the significance and use the receiver operating characteristic curve method to find the optimal cut-point for these significant features. After splitting 80% of the data into a training set and 20% data into a testing set, we use five machine learning models, k-nearest Neighbors (kNN), Naïve Bayes, Logistic regression, Artificial Neural Networks (ANNs) and supportive vector machine (SVM), to explore and validate the performance of US GLCM texture features in diagnosing salivary glandular tumors.

RESULTS

This study includes 49 female and 73 male patients, with a mean age of 53 years old, ranging from 21 to 93. We find that six GLCM texture features (contrast, inverse difference movement, entropy, dissimilarity, inverse difference and difference entropy) are significantly different between benign and malignant tumors (p < 0.05). In ML, the overall accuracy rates are 74.3% (95%CI: 59.8-88.8%), 94.3% (86.6-100%), 72% (54-89%), 84% (69.5-97.3%) and 73.5% (58.7-88.4%) for kNN, Naïve Bayes, Logistic regression, a one-node ANN and SVM, respectively.

CONCLUSIONS

US texture analysis with ML has potential as an objective and valuable tool to make a differential diagnosis between benign and malignant salivary gland tumors.

Comparative analysis of ultrasonographic fetal lung texture in twin and singleton fetuses

OBJECTIVES

Increased fetal lung heterogeneity has been associated with term fetal lungs in singleton gestations. The objective of this study was to determine if fetal lung heterogeneity index (HI) differs between twin and singleton fetuses in the late second and third trimesters.

METHODS

Prospective cohort study of women with singleton and twin gestations with medically-indicated ultrasound examinations at 24 weeks of gestation onward. Grayscale transverse fetal lung images were obtained at the level of the four-chamber heart. A region of interest was selected in each fetal lung image. Fetal lung HI was determined with MATLAB software using a dithering technique with ultrasound image pixels transformed into a binary map form from which a dynamic range value was determined. HI averages and standard deviations were generated for twin and singleton fetuses from 24 weeks gestation onward. Two sample t-tests were used to compare the mean HI at each gestational week between singleton and twin fetuses.

RESULTS

In total, 388 singleton and 478 twin images were analyzed. From 35 through 38 weeks of gestation a statistically significant divergence in mean HI was observed with higher means in singleton compared to twin fetuses. At 24 weeks of gestation there was a significantly higher HI in twin fetuses compared to singletons. No differences in fetal lung HI were observed between 25 and 34 weeks gestational age.

CONCLUSIONS

Differences in fetal lung HI were observed when comparing twin and singleton fetuses. Further investigation is required to determine the potential clinical significance of these findings.

Use of Virtual Touch Tissue Quantification Elastography Technique in Fetal Lung Maturation: A Preliminary Study

ABSTRACT

This study is an analysis of fetal lung stiffness by virtual touch tissue quantification (VTTQ) elastography to predict fetal lung maturation. Evaluation of fetal lungs was first performed in B mode, and fetal lungs were analyzed at 3 different periods at third trimester in each pregnant woman, at 28 to 31, 32 to 36, and 37 to 41 weeks. Fetal lung elastography was performed at regions with the least acoustic shadow and far from ribs and heart. Each fetal lung assessment were done by taking mean lung stiffness obtained by measuring stiffness of both left and right fetal lungs. T test analysis showed no significant difference in fetal lung stiffness between male and female fetuses among 3 gestational periods. Analysis of variance was performed to evaluate fetal lung stiffness of the fetuses at 3 different gestational periods (28-31, 32-36, and 37-41 weeks). This analysis showed significant difference ( P < 0.01). Duncan multiple comparison analysis did not show significant difference in fetal lung stiffness between 28 and 31 weeks and 32 and 36 weeks, whereas fetal lung stiffness of fetuses at 37 to 41 weeks were significantly greater ( P < 0.01). This study is first step to analyze fetal lung maturation noninvasively using VTTQ elastography technique by measuring fetal lung stiffness.

The effect of colorization on the fetal lung heterogeneity index

No previous studies have quantitatively assessed the effect of color tones on ultrasound texture analysis techniques. Our objective was to compare heterogeneity index (HI) between fetal lung images captured in grayscale and those same images after conversion to Ice and Sepia. Fetal lung images were obtained during medically indicated ultrasound examinations. We observed that HI is affected by the application of color tones to ultrasound images of the fetal lung. Therefore, for each type of biological tissue and color tone, determination of distinct HI nomograms and cut off points is recommended.

Fetal Growth and Osteogenesis Dynamics during Early Development in the Ovine Species

Increased knowledge of the developmental processes during gestation could provide valuable information on potential alterations in embryonic/fetal development. We examined the development of ovine conceptus between the 20th and 70th day of gestation with three convergent analyses: (1) uterus ultrasound examination and measurement (eco) of crown-rump length (CRL) and biparietal diameter (BPD) of the conceptus; (2) direct measurement (vivo) of CRL and BPD of the conceptus outside the uterus (3) osteo-cartilage dynamics during development by differential staining. No significant differences were observed between eco and vivo measurements for CRL and BPD in all examined concepti. CRL and BPD, instead, showed a significant positive linear correlation with gestational age. The study of osteogenesis dynamics has demonstrated a completely cartilaginous ovine fetus at up to 35 days of gestation. The ossification begins in the skull (40th day) and is almost complete between the 65th and the 70th of pregnancy. Our study highlighted that CRL and BPD are accurate parameters for gestational age estimation in the first part of sheep pregnancy and provides an overview of osteochondral temporal dynamics. Furthermore, tibia ossification is a valid parameter to estimate fetal age by ultrasound.

Artificial Intelligence in Quantitative Ultrasound Imaging: A Survey

Quantitative ultrasound (QUS) imaging is a safe, reliable, inexpensive, and real-time technique to extract physically descriptive parameters for assessing pathologies. Compared with other major imaging modalities such as computed tomography and magnetic resonance imaging, QUS suffers from several major drawbacks: poor image quality and inter- and intra-observer variability. Therefore, there is a great need to develop automated methods to improve the image quality of QUS. In recent years, there has been increasing interest in artificial intelligence (AI) applications in medical imaging, and a large number of research studies in AI in QUS have been conducted. The purpose of this review is to describe and categorize recent research into AI applications in QUS. We first introduce the AI workflow and then discuss the various AI applications in QUS. Finally, challenges and future potential AI applications in QUS are discussed.

Values and validity of fetal parameters by ultrasound and Doppler as markers of fetal lung maturity

Background

In this study, we combined ultrasound and Doppler parameters to conclude the most accurate and applicable method for fetal lung maturity assessment. Thepurpose is to reduce risk of neonatal respiratory distress by assessment of fetal lung maturity through ultrasound and Doppler instead of amniocentesis.

Results

A total of 120 women were included. The ultrasound examination was performed at a gestational age ranging between 30 and 37 weeks and delivery occurred within 72 h from ultrasound. After birth, 19 fetuses were diagnosed with RDS.

The ultrasound parameters assessed for fetal lung maturity showed varying sensitivity and specificity but presence of proximal tibial epiphyses showed the highest sensitivity (91%) and specificity (95%) followed by the presence of distal femoral epiphyses with 90% sensitivity and 84% specificity. The placental grade III maturity showed a sensitivity of 81% and 74% specificity followed by the presence of the amniotic fluid free-floating particles with an 83% sensitivity and 63% specificity, while echogenic fetal thalamus showed 77% and 79% for sensitivity and specificity, respectively, for prediction of fetal lung maturity. By using the fetal biometry, we could report that a BPD between 82.8 and 93.5 mm, AC between 295 and 322 mm, and FL (62.7-72.1 mm) correlated with mature fetal lungs. Regarding Doppler indices, increased resistive index of umbilical artery > 0.69 and of pulmonary artery > 0.78 correlated with neonatal RDS while fetuses with UtA RI values between 0.43 and 0.66 noticed in RDS vs (0.42-0.57) in the non-RDS. MCA RI values are also higher in RDS 0.77–0.88 vs 0.74–0.79 in non-RDS.

On the other hand, analysis of colon grading and fetal lung echogenicity in relation to liver showed no significant value in the assessment of fetal lung maturity.

Conclusion

Ultrasound and Doppler showed good predictive value and accuracy and is considered a valuable non-invasive procedure in assessment of fetal lung maturity. However, no single parameter alone could show a definitive sign

Artificial Intelligence in Obstetric Ultrasound: An Update and Future Applications

Artificial intelligence (AI) can support clinical decisions and provide quality assurance for images. Although ultrasonography is commonly used in the field of obstetrics and gynecology, the use of AI is still in a stage of infancy. Nevertheless, in repetitive ultrasound examinations, such as those involving automatic positioning and identification of fetal structures, prediction of gestational age (GA), and real-time image quality assurance, AI has great potential. To realize its application, it is necessary to promote interdisciplinary communication between AI developers and sonographers. In this review, we outlined the benefits of AI technology in obstetric ultrasound diagnosis by optimizing image acquisition, quantification, segmentation, and location identification, which can be helpful for obstetric ultrasound diagnosis in different periods of pregnancy.

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

OBJECTIVES

To analyze and compare, using ultrasound-based radiomics technology, fetal-lung texture in pregnancies affected by gestational diabetes mellitus (GDM) and/or pre-eclampsia (PE) and in normal pregnancies, overall and at different gestational ages.

METHODS

In this retrospective study, 430 high-throughput features per fetal-lung image were extracted from 548 fetal-lung ultrasound images (obtained at the level of the four-chamber view of the heart) in 548 pregnant women who delivered between July 2018 and August 2019 at the Obstetrics and Gynecology Hospital of Fudan University. Images had been obtained during ultrasound examinations between 28 and 41 weeks of gestation. The data were divided randomly into training set (80% of fetal-lung images) and independent test set (20% of images), and 20% of the images in the training set were then selected as the validation set. A standard machine-learning model based on ultrasound-based radiomics technology was created using features of fetal-lung texture extracted from the images, and a regression model was used to evaluate the relationship between lung-texture features, GDM and/or PE and gestational age.

RESULTS

Of the 548 pregnancies included, 108 were affected by GDM alone, 71 by PE alone and 25 by both GDM and PE, and 344 were normal. The overall performance of the GDM and PE prediction model was superior to that of the gestational-age prediction model, with an area under the receiver-operating-characteristics curve of 0.95-0.99, sensitivity of 78.8-97.1% in the validation set and 74.5-91.3% in the independent test set, specificity of 79.8-94.3% in the validation set and 75.7-88.4% in the independent test set and accuracy of 81.0-95.3% in the validation set and 80.6-86.4% in the independent test set.

CONCLUSIONS

Using ultrasound-based radiomics technology, fetal lungs from pregnancies grouped according to whether they were affected by GDM and/or PE could be distinguished from each other and from fetal lungs of normal pregnancies, and lungs from pregnancies at different gestational ages could be distinguished. These findings support further research to explore the use of this non-invasive technology to predict neonatal respiratory complications in women with PE, GDM or their combination. © 2020 International Society of Ultrasound in Obstetrics and Gynecology.

Fetal Lung Echo Texture in Pregnancies at Risk for Pulmonary Hypoplasia

Pulmonary hypoplasia is associated with severe respiratory distress immediately after birth and frequently leads to neonatal death. In this study, we compared the fetal lung echo texture in pregnancies at high and low risk for pulmonary hypoplasia. Ultrasonic tissue heterogeneity was determined by a dynamic range calculation. This quantification uses a dithering technique based on the Floyd-Steinberg algorithm, in which the pixels are transformed into a binary map. Pregnancies at high risk for pulmonary hypoplasia showed decreased fetal lung heterogeneity on ultrasound imaging. This image-processing technique may allow improved risk stratification, patient counseling, and treatment approaches for pulmonary hypoplasia.

The Classification and Segmentation of Fetal Anatomies Ultrasound Image: A Survey

Ultrasound imaging processing technology has been used in obstetric observation of the fetus and diagnosis of fetal diseases for more than half a century. It contains certain advantages and unique challenges which has been developed rapidly. From the perspective of ultrasound image analysis, at the very beginning, it is essential to determine fetal survival, gestational age and so on. Currently, the fetal anatomies ultrasound image analysis approaches have been studies and it has become an indispensable diagnostic tool for diagnosing fetal abnormalities, in order to gain more insight into the ongoing development of the fetus. Presently, it is the time to review previous approaches systematically in this field and to predict the directions of the future. Thus, this article reviews state-of-art approaches with the basic ideas, theories, pros and cons of ultrasound image technique for whole fetus with other anatomies. First of all, it summarizes the current pending problems and introduces the popular image processing methods, such as classification, segmentation etc. After that, the advantages and disadvantages in existing approaches as well as new research ideas are briefly discussed. Finally, the challenges and future trend are discussed.

Evaluation of an improved tool for non-invasive prediction of neonatal respiratory morbidity based on fully automated fetal lung ultrasound analysis

The objective of this study was to evaluate the performance of a new version of quantusFLM®, a software tool for prediction of neonatal respiratory morbidity (NRM) by ultrasound, which incorporates a fully automated fetal lung delineation based on Deep Learning techniques. A set of 790 fetal lung ultrasound images obtained at 24 + 0–38 + 6 weeks’ gestation was evaluated. Perinatal outcomes and the occurrence of NRM were recorded. quantusFLM® version 3.0 was applied to all images to automatically delineate the fetal lung and predict NRM risk. The test was compared with the same technology but using a manual delineation of the fetal lung, and with a scenario where only gestational age was available. The software predicted NRM with a sensitivity, specificity, and positive and negative predictive value of 71.0%, 94.7%, 67.9%, and 95.4%, respectively, with an accuracy of 91.5%. The accuracy for predicting NRM obtained with the same texture analysis but using a manual delineation of the lung was 90.3%, and using only gestational age was 75.6%. To sum up, automated and non-invasive software predicted NRM with a performance similar to that reported for tests based on amniotic fluid analysis and much greater than that of gestational age alone.

Renal sonographic changes in heterogeneity index and echogenicity in children with hypertension: a novel assessment

The kidneys are thought to contribute to the pathogenesis of primary hypertension, but hypertension is also known to cause target organ damage in the kidney. Noninvasive methods to capture possible changes in the kidney related to hypertension are limited. A new program that has been used to quantify the heterogeneity and percent echogenicity in renal ultrasound images was implemented to assess patients with hypertension. Children and adolescents <21 years with primary hypertension diagnosed by ambulatory blood pressure monitoring were compared with normotensive age- and sex-matched controls. Renal ultrasound images were evaluated by a technique that measured pixels of gray-scale images and transformed them into a binary map, which was converted to a heterogeneity index (HI) and percent echogenicity score. This study included 99 children with hypertension and 99 control subjects. Body mass index (BMI) was greater in the hypertension group. Average HI for hypertension was significantly higher than in controls (1.37 ± 0.19 vs. 1.2 ± 0.23, P = .001), while echogenicity scores were not different (26.6 ± 8.9 vs. 25.9 ± 10, P = .8). In regression analysis adjusting for BMI z-score and race, hypertension was associated with greater HI compared with controls (β = 0.11, 95% confidence interval 0.03-0.18, P = .005). In a model adjusted for age, sex, and BMI z-score in the hypertension group only, no ambulatory blood pressure monitoring measures were associated with HI or echogenicity scores (P > .05).HI was significantly greater in the hypertension group compared with normotensive controls. HI may be a novel method to detect changes in the kidney related to hypertension.

Quantitative Ultrasound Analysis of Proximal and Distal Cervical Tissue Echogenicity in Premature Cervical Remodeling

OBJECTIVES

To determine whether a novel, noninvasive quantitative ultrasound (US) technique can detect differences in proximal and distal cervical tissue echogenicity in women with and without a shortened cervical length (CL).

METHODS

We conducted a retrospective case-control study of singleton pregnancies at 16 to 26 weeks' gestation in which a transvaginal US examination was performed to measure CL from 2013 to 2015. Initial CLs in cases and controls were less than 2.5 cm and 2.5 cm or greater, respectively. For each US image, a region of interest was selected in the proximal and distal cervical stroma, in both the anterior and posterior cervical lips. The Floyd-Steinberg dithering algorithm transformed grayscale pixels in each region of interest into a binary map. A histogram tabulated the number of black and white pixels, allowing determination of the percent echogenicity. The difference in the percent echogenicity was calculated by subtracting the distal cervical echogenicity (average of anterior and posterior lips) from the proximal cervical echogenicity (average of anterior and posterior lips).

RESULTS

Ultrasound images from 177 women were analyzed. There was a difference in the percent echogenicity (P < .0001) when comparing women with a short cervix (mean ± SD, 9.8 ± 10.1; n = 102) to women with a normal CL (17.2 ± 9.5; n = 75). Differences were attributable to changes in proximal (P < .008) rather than distal cervical echogenicity. Regardless of CL, the proximal cervix was more echogenic than the distal cervix.

CONCLUSIONS

A quantitative US analysis of cervical tissue can detect differences in echogenicity between the proximal and distal cervix in the second trimester. Proximal cervical echogenicity is lower with CL of less than 2.5 cm compared to a normal CL.