Quantitative Ultrasound TextureAnalysis for Differentiating PretermFrom Term Fetal LungsSleiman R. Ghorayeb, PhD, Luis A. Bracero, MD, Matthew J. Blitz, MD, MBA, Zara Rahman, BS,Martin L. Lesser, PhD

Objectives—To differentiate preterm (<37 weeks’ gestation) from term (�37weeks’ gestation) fetal lungs by using quantitative texture analysis of ultrasoundimages.

Methods—This study retrospectively evaluated singleton gestations with valid dat-ing at 20 weeks’ gestational age (GA) or later between January 2015 and December2015. Images were obtained from Voluson E8 ultrasound systems (GE Healthcare,Milwaukee, WI). A region of interest was selected in each fetal lung image at thelevel of the 4 heart chambers from an area that appeared most representative of theoverall lung tissue and had the least shadow. Ultrasonic tissue heterogeneity (hetero-geneity index) based on dynamic range calculation was determined for all lungimages. This quantification was performed with a custom-made software programthat used a dithering technique based on the Floyd-Steinberg algorithm, in whichthe pixels are transformed into a binary map. Regression analysis was used to deter-mine the correlation and functional association between the heterogeneity indexand GA. A receiver operating characteristic curve was used to identify the optimalheterogeneity index cutoff point for differentiating preterm from term fetal lungs.

Results—A total of 425 fetal lung ultrasound images (313 preterm and 112 term)were analyzed. Quantitative texture analysis predicted GA with sensitivity and speci-ficity of 87.9% and 92.0%, respectively, based on the optimal receiver operatingcharacteristic cutoff point.

Conclusions—Quantitative ultrasound texture analysis of fetal lung tissue can differ-entiate preterm fetal lungs from term fetal lungs. Our data suggest that decreasedfetal lung heterogeneity on ultrasound imaging is associated with preterm fetuses.

Key Words—fetal lung tissue; heterogeneity index; obstetric ultrasound;quantitative ultrasound

L ung immaturity remains the most common cause of morbidityand mortality in preterm and early term neonates.1 Althoughgestational age (GA) is the best predictor of lung maturity,

respiratory distress syndrome and transient tachypnea of the neonateare not restricted to very preterm (<34 weeks’ gestation) births.Late preterm (34–36 weeks’ gestation) and early term (37–38 weeks’gestation) neonates have significantly higher rates of these complica-tions than neonates born at or beyond 39 weeks’ gestation.2,3 Thus,identification of fetuses at risk for respiratory morbidity after birthdue to lung prematurity or hypoplasia remains very important in thepractice of obstetrics. The decision to proceed with or delay delivery

Received June 29, 2016, from the School of Engineer-ing and Applied Sciences, Ultrasound Research Labo-ratory, Hofstra University, Hempstead, New YorkUSA (S.R.G, Z.R.); Departments of Radiology andMolecular Medicine, Hofstra Northwell School ofMedicine, Hempstead, New York USA (S.R.G.);Center for Immunology and Inflammation, FeinsteinInstitute for Medical Research, Northwell Health,Manhasset, New York USA (S.R.G.); Division ofMaternal-Fetal Medicine, Department of Obstetricsand Gynecology, Hofstra Northwell School of Medi-cine, Southside Hospital, Bay Shore, New York USA(L.A.B.); Division of Maternal-Fetal Medicine,Department of Obstetrics and Gynecology, HofstraNorthwell School of Medicine, North Shore UniversityHospital, Manhasset, New York USA (M.J.B.); andBiostatistics Unit, Feinstein Institute for MedicalResearch, Hofstra Northwell School of Medicine,Manhasset, New York, USA (M.L.L.). Manuscriptaccepted for publication September 29, 2016.

Some of this work was presented at the 2016Third International Conference on Biomedical Engi-neering and Systems; August 16–17, 2016; Budapest,Hungary. The statistical analysis was partially doneat the School of Engineering and Applied Sciences,Ultrasound Research Laboratory, and then continuedin the Biostatistics Unit, Feinstein Institute for MedicalResearch, Hofstra Northwell School of Medicine. Theultrasound images in this study were collected at theCenter for Maternal-Fetal Health at Suffolk (Smith-town, NY) and the Southside Hospital AntenatalTesting Unit (Bay Shore, NY), and the image analysisportion of this work was performed at the School ofEngineering and Applied Sciences, Ultrasound Re-search Laboratory, Hofstra University.

Address correspondence to Sleiman R.Ghorayeb, PhD, School of Engineering and AppliedSciences, Hofstra University, 104 Weed Hall,Hempstead, NY 11549-0001 USA.

E-mail: sghorayeb@northwell.edu, sleiman.r.ghorayeb@hofstra.edu

AbbreviationsGA, gestational age; ROI, region of interest; ROC,receiver operating characteristic

doi:10.7863/ultra.16.06069

VC 2017 by the American Institute of Ultrasound in Medicine | J Ultrasound Med 2017; 36:1437–1443 | 0278-4297 | www.aium.org

ORIGINAL RESEARCH

is often dependent on the ability to properly assess fetallung maturity and the presence or absence of lunghypoplasia.

Rapid, precise, and diagnostically sensitive tests oflung maturity are now available, but unfortunately,they remain invasive, have low specificity, and remainpoor predictors of fetal lung immaturity.4,5 The mostcommonly available tests assess the lamellar bodycount, lecithin-to-sphingomyelin ratio, and presenceor absence of phosphatidylglycerol. All involve testingamniotic fluid via amniocentesis to provide an indirectassessment of the likelihood of lung maturity.Although the risks associated with amniocentesis inthe third trimester are low, complications such asfetal injury, preterm labor, placental abruption, mater-nal sepsis, fetal heart rate abnormalities, and fetal-maternal hemorrhage have been reported.6

Noninvasive ultrasound methods capable of distin-guishing between mature and immature fetal lungswould be of important clinical value. A number of suchultrasound techniques have been studied with varyingdegrees of success. Some have focused on direct assess-ment of the fetal lung by measurement of fetal lung vol-umes. Others have relied on indirect associations withfetal lung maturity, including placental grading, measure-ment of epiphysis ossification centers, and pulmonaryartery Doppler studies.7–9 Furthermore, accurately pre-dicting lethal pulmonary hypoplasia in the fetus has alsoproven difficult by ultrasound imaging.10

Quantitative ultrasound texture analysis of the fetallung has been proposed as a promising noninvasivemethod to predict fetal lung maturity, fetal lung hypopla-sia, and neonatal respiratory morbidity.11–14 This tech-nique uses standard fetal lung images, which are easilyobtained by sonographers during routine ultrasoundexaminations. Additional information can then be ex-tracted from these images by applying quantitative proc-essing methods that characterize the tissue. Previousstudies investigating this approach have used a variety ofmethods. Some have been small studies focused on feasi-bility, reliability, and validation of technique.13 Othershave been large investigations focused on the predictionof clinical events.11 To date, it remains uncertain whe-ther alternative texture analysis approaches can improveon the techniques previously demonstrated.

Our objective was to differentiate preterm (<37weeks’ gestation) from term (�37 weeks’ gestation)fetal lungs by using a new quantitative ultrasound texture

analysis method. This study was a preliminary applica-tion of a novel technique, which transforms the pixels ofan ultrasound image into a binary map by employing aFloyd-Steinberg dithering algorithm. Unlike similar priorstudies, the technique demonstrated here will be capableof using a smaller region of interest (ROI) within thefetal lung and will yield an index value rather than adichotomous (yes/no) response.

Materials and Methods

This retrospective cohort study included women withsingleton gestations who had routine pregnancy ultra-sound examinations at 20 weeks’ GA or later, for a vari-ety of indications, at the Center for Maternal-FetalHealth at Suffolk (Smithtown, NY) and the SouthsideHospital Antenatal Testing Unit (Bay Shore, NY)between January 2015 and December 2015. Ultrasoundexaminations with a 4-chamber view of the fetal heartwere sequentially selected from our ultrasound database.Protected health information was removed from allextracted ultrasound images before the texture analysis.All pregnancies included in the study had valid dating,based on the last menstrual period and measurement ofthe crown-rump length during the first trimester. Exclu-sion criteria were multiple gestations, fetal malforma-tions, and pregnancies with screening tests suggestive offetal aneuploidy. Patients were not excluded for anymaternal medical conditions. The Northwell Health Sys-tem Human Research Protection Program and Institu-tional Review Board approved the study protocol.

Image Acquisition and AnalysisFetal lung image acquisition was achieved by using atransverse view of the fetal thorax at the level of the 4-chamber view. All images were obtained with GE Volu-son E8 ultrasound systems (GE Healthcare, Milwaukee,WI) for uniformity. The ultrasound machines were allequipped with convex array transducers with a fre-quency range of 3 to 7.5 MHz. Ultrasound examinationswere performed by experienced sonographers, and amaternal-fetal medicine attending physician reviewedtheir findings. Equipment settings were adjusted at thesonographer’s discretion to obtain the optimal imagequality. The only information retained with the ultra-sound images was the GA at the time of the ultrasoundexamination.

Ghorayeb et al—Fetal Lung Texture Analysis With Ultrasound

1438 J Ultrasound Med 2017; 36:1437–1443

Ultrasound images were digitally collected in theoriginal Digital Imaging and Communications in Medi-cine format and then analyzed in a custom-made pro-gram with a graphical user interface tool developed inMATLAB (version 8.5.0.197613 R2015a; The Math-Works, Inc, Natick, MA). The software uses a ditheringtechnique based on the Floyd-Steinberg algorithm, inwhich the pixels of an ultrasound image are transformedinto a binary map from which a dynamic range value isdetermined.15–17 This program, in short, converts animage to gray scale and rounds the intensity of each pixelto its nearest extreme, black or white. It then calculatesthe percentage of white pixels with respect to the totalnumber of pixels in a selected ROI. This method is thenbased on the ratio of the maximum to minimum of thegrayscale pixel values. Therefore, adjustment of the gainby the sonographer will not affect the result. For eachfetal lung, an ROI was manually selected by a single phy-sician, carefully ensuring that only lung tissue wasdelineated. The ROI was selected from the area of thelung that had the least shadow. Each ROI was a square,which was further divided into at least 25 sub-ROIs(5 3 5) with each sub-ROI containing 100 pixels, for aminimum of 2500 pixels. Larger areas were chosenwhen the images were of good quality, and smaller areaswere selected when there was shadowing. For example,if the selected ROI was larger and had 10 3 10 sub-ROIs, then a total of 10,000 pixels would be analyzed. Inall cases, the area most representative (ie, most uniform,consistent, and homogeneous) of the overall lung tissuewas chosen. A quantification of ultrasonic tissue hetero-geneity (heterogeneity index) was then determined bycalculating the dynamic range for each ROI. An ultra-sound image representing the major features of the fetallung, showing the selected ROI and the sub-ROIs withinthe ROI, is shown in Figure 1.

For each ultrasound image, the heterogeneity indi-ces of the left and right lungs were averaged, and then adecision matrix was used to classify the heterogeneityindex. All lungs belonging to a GA if 37 weeks or later(term fetuses) were taken to be the actual (true) nega-tive value, whereas those with GA earlier than 37 weeks(preterm fetuses) were taken to be the actual (true) pos-itive value (Table 1).

Statistical AnalysesIt should first be emphasized that the primary objectivesof the statistical analyses were to derive an optimal

“cutoff point” for the heterogeneity index as used to pre-dict a preterm fetal lung and to evaluate the sensitivityand specificity by using that cutoff point rule. To derivean initial cutoff point rule, the mean and standard devia-tion of the heterogeneity index for the GA group of 37weeks or later were calculated, and a “term referencerange” was calculated as mean 6 2 SDs. If the heteroge-neity index was higher than the lower limit of the termreference range (ie, heterogeneity index >mean – 2SDs), then the image was classified as belonging to GAof 37 weeks or later. Conversely, if the heterogeneityindex was less than mean – 2 SDs, then the image wasclassified as belonging to GA earlier than 37 weeks. Sen-sitivity and specificity were then computed directly fromthe resulting 2 3 2 table (Table 1) in the usual way.Next, logistic regression (SAS version 9.4; SAS InstituteInc, Cary, NC) was used to estimate the probability of a

Figure 1. Ultrasound image of the transverse view of a fetal thoraxtaken at the level of the 4-chamber view. In this case, the ROI hasbeen divided into 6 3 7 sub-ROIs, each of which contains 100 pixels,thus yielding a total pixel count of 4200.

Table 1. Calculation of Sensitivity and Specificity Depending onthe Heterogeneity Index (Based on an Optimal ROC HeterogeneityIndex Cutoff of <1.82)

Group <37 wk GA �37 wk GA Total

Low heterogeneityindex

275 (true-positive)Sensitivity: 87.9%

9 (false-positive)

284

High heterogeneityindex

38 (false-negative)

103 (true-negative)

Specificity:92.0%

141

Total 313 112 425

Ghorayeb et al—Fetal Lung Texture Analysis With Ultrasound

J Ultrasound Med 2017; 36:1437–1443 1439

preterm lung and the associated receiver operating char-acteristic (ROC) curve. An “optimal” cutoff point forthe index was derived as the point yielding sensitivityand 1 – specificity closest to the point (0, 1) on theROC curve. Regression analysis (SAS) was used todetermine the correlation and (linear) functional associ-ation between the index and GA.

Before the study was conducted, the following sam-ple size calculation was performed: For estimating pre-sumed sensitivity of at least 75%, a sample size of 300preterm fetuses was required to construct a 95% confi-dence interval with error of at most 65%. A sample sizeof 100 term fetuses yields an error of at most 68.6%.

It is important to stress that the study objective wasnot to determine whether a fetal lung is “premature” or“underdeveloped.” Rather, the objective was to deter-mine whether the fetal lung belonged to a fetus who wasearlier than 37 weeks or 37 weeks or later. Accordingly,positive and negative predictive values, whereas calcu-lated, must be interpreted appropriately. For example, ifthe test is applied to a fetus of, say, 35 weeks, the GA isperfectly known to be preterm, so the prediction ofwhether the lung is preterm is moot. However, if theresult of the heterogeneity index test is “negative” (ie, itis predicting lungs from a term fetus), then we mightconclude that the prediction of the GA group is inconsis-tent with the known group.

Results

Ultrasound images from 427 patients were collected.Two patients were removed before the analyses becauseof poor image quality. A total of 425 fetal lung images(313 preterm and 112 term) were analyzed. The meanheterogeneity index and SD for GA of 37 weeks or laterwere 2.33 and 0.51, respectively; and those for GA ear-lier than 37 weeks were 1.51 and 0.26 (Figure 2). Themeans were significantly different (P< .0001, t test).

For the group with GA of 37 weeks or later, thelower limit of the term reference range was computed as1.13, thus classifying an image as earlier than 37 weeks ifthe heterogeneity index was less than 1.13. With the useof this cutoff point for classification, the computed sensi-tivity was 23.6% (74 of 313), and specificity was 96.4%(108 of 112).

The ROC curve analysis (Figure 3) resulted in areaunder the curve of 0.935. The optimal cutoff point wasfound to be “classify as preterm” if the heterogeneity

index was less than 1.82. Using this cutoff point resultedin sensitivity of 87.9% (275 of 313) and specificity of92.0% (103 of 112). Table 1 shows the frequency distri-bution of the heterogeneity index cutoff rule for eachgroup.

Figure 2. Side-by-side box plots comparing GA earlier than 37 weeksand 37 weeks and later. The middle horizontal line is the median; thediamond is the mean; and lower and upper box edges are 25th and75th percentiles, respectively. HI indicates heterogeneity index.

Figure 3. Receiver operating characteristic curve showing the areaunder the curve that corresponds to the accuracy of theheterogeneity index (HI) as a predictor of GA earlier than 37 weeks.The optimal heterogeneity index cutoff point of less than 1.82 (topredict GA< 37 weeks) corresponds to the point with sensitivity of87.9% and specificity of 92.0%.

Ghorayeb et al—Fetal Lung Texture Analysis With Ultrasound

1440 J Ultrasound Med 2017; 36:1437–1443

Regression analysis of the heterogeneity index as afunction of GA yielded a slope of 0.06 (P< .0001) andR2 of 33%. Thus, for every 1 week of GA, the heteroge-neity index increased by 0.06 units. Figure 4 shows ascatterplot of the mean heterogeneity index versus GA.The smoothed spline curve suggests nonlinearity as GAexceeds 34 weeks.

Discussion

Our results indicate that quantitative ultrasound textureanalysis can indicate changes that occur in fetal lung tis-sue as gestation advances. We found that decreased tis-sue heterogeneity on ultrasound imaging is associatedwith preterm fetal lungs, and increased tissue heteroge-neity is associated with term fetal lungs. It is importantto note that these findings do not indicate that suchlungs are mature or immature in terms of their physio-logic development. However, given that there is a knowncorrelation between preterm lungs and premature pul-monary function, future studies should evaluate whetherthis technique can also differentiate between mature andimmature fetal lungs and thus predict neonatal respira-tory morbidity.

Using standard statistical ROC curve methods forfinding an “optimal” cutoff point to differentiate a pre-term lung from a term lung, a cutoff point of 1.82 wasderived (heterogeneity index< 1.82 5 preterm; hetero-geneity index� 1.82 5 term), yielding sensitivity andspecificity of 87.9% and 92.0%, respectively. These oper-ating characteristics, along with an area under the ROC

curve of 93.5%, suggest that the heterogeneity index isan excellent discriminator.

During the saccular stage of lung development,which begins at approximately 24 weeks’ gestation, thereis progressive thinning of the epithelium, terminal saccu-lar formation, and the start of surfactant production.The transition to the alveolar stage of lung developmentis not sharply defined. Alveoli can be visualized as earlyas 29 weeks’ gestation, and lung volume increases 4-foldbetween this time and term.18 This process typicallyaccelerates at approximately 36 to 38 weeks’ gestationand continues through infancy.19,20 During this period,alveolar formation and septation occur, allowing expan-sion of air spaces. Presumably, this process is the changein the fetal lung structure that we are detecting.

A variety of techniques have been used in previousstudies investigating quantitative ultrasound for assessingfetal lung maturity. Prakash et al21 developed a methodin which textural features of both the fetal lung and liverwere computed, and the ratio of lung-to-liver values wasevaluated to predict lung maturity, using GA as a surro-gate. This technique had reported accuracy of 73% to96%. Tekesin et al22 evaluated the mean gray value ofthe fetal lung and demonstrated a characteristic patternas gestation advanced, which corresponded to fetal lungdevelopment, but the only significant differences werenoted before 32 weeks’ gestation. Serizawa and Maeda23

evaluated the gray-level histogram width of the fetal lungand liver to predict fetal lung immaturity, with sensitivityof 96% and a specificity of 72%. More recently, quanti-fication of lung texture has been performed with“automated quantitative ultrasound analysis” software,which is not affected by changes in illumination anddoes not rely on tissue references such as the liver or thedirect gray level from the image.12,13 This methodshowed a strong correlation (R 5 0.98) with GA13 andwas able to predict fetal lung maturity, as confirmedby amniocentesis, with sensitivity of 95.1%, specificityof 85.7%, and accuracy of 90.3%.12 This group hassince developed a newer software platform known as“quantitative ultrasound fetal lung maturity analysis,”which combines various image texture extractors andmachine-learning algorithms to predict neonatal respira-tory morbidity.11 This method was prospectively andblindly validated, with results that were comparable withcurrent tests using amniotic fluid, but the study waslimited by a relatively small sample size from a singleinstitution.

Figure 4. Scatterplot of mean heterogeneity index (HI) versus GA.The smooth line shows the general trend.

Ghorayeb et al—Fetal Lung Texture Analysis With Ultrasound

J Ultrasound Med 2017; 36:1437–1443 1441

Our study had several strengths. With more than400 images, it was sufficiently powered to detect a differ-ence in the heterogeneity index between preterm andterm fetal lungs. In addition, the technique investigatedin this study uses standard fetal lung images to obtainadditional clinically relevant information and requires nospecialized nonstandard ultrasound procedures, whichwould likely require extensive training and standardiza-tion to implement into practice. Moreover, this methoddoes not use other tissue references such as the liver anddoes not use the direct gray level of the ultrasoundimage. Another strength was that we did not considerthe maternal and obstetric conditions, if any, associatedwith each fetal lung image.

Our study also had several limitations, including thelack of outcome data and the fact that there was no cor-relation with fetal lung immaturity or maturity; that is,fetuses at earlier 37 weeks’ GA may have mature lungs,and those at 37 weeks’ GA or later may have immaturelungs. Maternal and fetal factors associated with anincreased or decreased risk of neonatal respiratory mor-bidity were unknown for this study population. Specifi-cally, prematurity, low birth weight, advanced maternalage, male sex, cesarean delivery before the onset of labor,and maternal conditions such as diabetes, hypertension,and nonhypertensive renal disease are associated with anincreased risk of respiratory distress syndrome.24 A dec-reased risk of respiratory distress syndrome is associatedwith antenatal corticosteroid administration, preeclamp-sia, fetal growth restriction, and maternal smoking. Fur-thermore, although there may be a degree of subjectivityor arbitrariness associated with manual selection of theROI in each ultrasound image, the technique used toquantify the heterogeneity index indicates detects differ-ences in soft tissue heterogeneity that are not perceptibleto the naked eye, so any concerns about bias should bemitigated. In addition, our technique demonstrates thatthe ROI need not delineate the largest possible area ofthe lung, as some prior studies have done.

Our findings support further investigation of quanti-tative ultrasound texture analysis for noninvasive assess-ment of fetal lung tissue immaturity. If such a techniqueis implemented into clinical practice, it may have a sub-stantial impact on neonatal outcomes, given the growingnumber of elective deliveries, late preterm births, andalso early term births. Despite interventions such asantenatal corticosteroids and postnatal surfactants, lungimmaturity remains a major cause of neonatal morbidity

and mortality. Although some deliveries should occureven in the absence of documented fetal lung tissuematurity, there are many clinical scenarios in which mul-tiple maternal and fetal factors are considered, and deliv-ery may be thought of as a reasonable option but alsoone that could be postponed if there were evidence offetal lung tissue immaturity. The determination of lungtissue maturity by amniocentesis is not only invasive butalso time-consuming, labor intensive, and expensive.Therefore, a noninvasive ultrasound method, as des-cribed in this study, may be of considerable clinical utilityin such cases. Of the 284 images that predicted GA ear-lier than 37 weeks (ie, heterogeneity index< 1.82), only9 (3.2%) were incorrect (ie, false-positive rate, 3.2%). Ofthe 141 images that predicted GA of 37 weeks or later(ie, heterogeneity index � 1.82), the false-negative ratewas 27.0%. Since the false-positive and false-negativerates depend on sensitivity, specificity, as well as preva-lence, these rates will vary among institutions as a func-tion of the “prevalence” of scans that are performedearlier than 37 weeks at a given institution or practice.Accordingly, the Bayes rule would need to be used torecompute the false-positive and false-negative rates. Wespeculate that these preterm fetuses (<37 weeks) with ahigh heterogeneity index may have had mature fetallungs, whereas the term fetuses (�37 weeks) with a lowheterogeneity index may have had immature lungs.

Future studies will evaluate this technique in a pro-spective multicenter manner and correlate the heteroge-neity index of the fetal lung with obstetric outcomes,including neonatal respiratory morbidity. Although thistechnique should function independent of the ultra-sound machine from which the image is derived, thisability must also be confirmed.

References

1. Robertson PA, Sniderman SH, Laros RK Jr, et al. Neonatal morbidity

according to gestational age and birth weight from five tertiary care

centers in the United States, 1983 through 1986. Am J Obstet Gynecol

1992; 166:1629–1645.

2. Spong CY, Mercer BM, D’Alton M, Kilpatrick S, Blackwell S, Saade

G. Timing of indicated late-preterm and early-term birth. Obstet

Gynecol 2011; 118:323–333.

3. Teune MJ, Bakhuizen S, Gyamfi Bannerman C, et al. A systematic

review of severe morbidity in infants born late preterm. Am J Obstet

Gynecol 2011; 205:374.e1–374.e9.

Ghorayeb et al—Fetal Lung Texture Analysis With Ultrasound

1442 J Ultrasound Med 2017; 36:1437–1443

4. Besnard AE, Wirjosoekarto SA, Broeze KA, Opmeer BC, Mol BW.

Lecithin/sphingomyelin ratio and lamellar body count for fetal lung

maturity: a meta-analysis. Eur J Obstet Gynecol Reprod Biol 2013; 169:

177–183.

5. Grenache DG, Gronowski AM. Fetal lung maturity. Clin Biochem

2006; 39:1–10.

6. Varner S, Sherman C, Lewis D, et al. Amniocentesis for fetal lung

maturity: will it become obsolete? Rev Obstet Gynecol 2013; 6:126–

134.

7. Schenone MH, Samson JE, Jenkins L, Suhag A, Mari G. Predicting

fetal lung maturity using the fetal pulmonary artery Doppler wave

acceleration/ejection time ratio. Fetal Diagn Ther 2014; 36:208–214.

8. Shah YG, Graham D. Relationship of placental grade to fetal pulmo-

nary maturity and respiratory distress syndrome. Am J Perinatol 1986;

3:53–55.

9. Tabsh KM. Correlation of ultrasonic epiphyseal centers and the

lecithin:sphingomyelin ratio. Obstet Gynecol 1984; 64:92–96.

10. Vergani P. Prenatal diagnosis of pulmonary hypoplasia. Curr Opin

Obstet Gynecol 2012; 24:89–94.

11. Bonet-Carne E, Palacio M, Cobo T, et al. Quantitative ultrasound tex-

ture analysis of fetal lungs to predict neonatal respiratory morbidity.

Ultrasound Obstet Gynecol 2015; 45:427–433.

12. Palacio M, Cobo T, Martinez-Terr�on M, et al. Performance of an

automatic quantitative ultrasound analysis of the fetal lung to predict

fetal lung maturity. Am J Obstet Gynecol 2012; 207:504.e1–504.e5.

13. Cobo T, Bonet-Carne E, Martinez-Terr�on M, et al. Feasibility and

reproducibility of fetal lung texture analysis by automatic quantitative

ultrasound analysis and correlation with gestational age. Fetal Diagn

Ther 2012; 31:230–236.

14. Beck AP, Araujo J�unior E, Leslie AT, Camano L, Moron AF.

Assessment of fetal lung maturity by ultrasound: objective study using

gray-scale histogram. J Matern Fetal Neonatal Med 2015; 28:617–622.

15. Ghorayeb SR, Levin A, Ast M, Schwartz JA, Grande DA. Sonographic

evaluation of knee cartilage defects implanted with preconditioned

scaffolds. J Ultrasound Med 2014; 33:1241–1253.

16. Ghorayeb SR, Rooney DM. Ultrasonic evaluation of bone quality in

cadaver ilia. Ann Biomed Eng 2013; 41:939–951.

17. Ghorayeb SR, Shah NV, Edobor-Osula F, et al. Ultrasonic assessment

of extracellular matrix content in healing Achilles tendon. IEEE Trans

Ultrason Ferroelectr Freq Control 2012; 59:694–702.

18. Hislop AA, Wigglesworth JS, Desai R. Alveolar development in the

human fetus and infant. Early Hum Dev 1986; 13:1–11.

19. Langston C, Kida K, Reed M, Thurlbeck WM. Human lung growth

in late gestation and in the neonate. Am Rev Respir Dis 1984; 129:

607–613.

20. Thurlbeck WM. Postnatal growth and development of the lung. Am

Rev Respir Dis 1975; 111:803–844.

21. Prakash KN, Ramakrishnan AG, Suresh S, Chow TW. An investiga-

tion into the feasibility of fetal lung maturity prediction using statistical

textural features. Ultrason Imaging 2001; 23:39–54.

22. Tekesin I, Anderer G, Hellmeyer L, Stein W, K€uhnert M, Schmidt S.

Assessment of fetal lung development by quantitative ultrasonic tissue

characterization: a methodical study. Prenat Diagn 2004; 24:671–676.

23. Serizawa M, Maeda K. Noninvasive fetal lung maturity prediction

based on ultrasonic gray level histogram width. Ultrasound Med Biol

2010; 36:1998–2003.

24. Luo G, Norwitz ER. Revisiting amniocentesis for fetal lung maturity

after 36 weeks’ gestation. Rev Obstet Gynecol 2008; 1:61–68.

Ghorayeb et al—Fetal Lung Texture Analysis With Ultrasound

J Ultrasound Med 2017; 36:1437–1443 1443

本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP

图书馆。

图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具