Share this content in WeChat
Clinical Article
A preliminary study on diagnostic model of placenta implantation based on magnetic resonance image feature machine learning
WANG Yingchao  HUANG Gang  BA Zhixia  XUE Lian  HUANG Baosheng  XIA Dongzhou 

WANG Y C, HUANG G, BA Z X, et al. A preliminary study on diagnostic model of placenta implantation based on magnetic resonance image feature machine learning[J]. Chin J Magn Reson Imaging, 2023, 14(8): 94-99, 107. DOI:10.12015/issn.1674-8034.2023.08.015.

[Abstract] Objective To explore the diagnostic value of machine learning model based on MRI T2WI imaging features for placenta accrete spectrum disorders (PAS).Materials and Methods The imaging data of 130 patients who underwent MRI examination and later caesarean section due to suspected placenta accretion were retrospectively analyzed. According to the postoperative results of caesarean section, MRI T2WI images were used to extract the imaging features of the layers with and without placenta accretion. The data were divided into a training set (n=91) and a validation set (n=39) by stratified sampling in a ratio of 7∶3. Five machine learning methods were adopted: logistic regression (LR), support vector machine (SVM), random forest (RF), decision tree (DT) and K nearest neighbor (KNN) for modeling and classification diagnosis. The hyperparameters of the machine learning model were determined by five-fold cross-validation. Receiver operating characteristic (ROC) curve was adopted to evaluate the prediction efficiency of the model, calculated the area under the curve (AUC), accuracy, sensitivity and specificity, and verified each model in the verification set. In addition, in addition to comparing the diagnostic effectiveness of different machine learning models with that of imaging diagnostic doctors, calibration curves were used to analyze the model efficacy and decision curve analysis (DCA) was used to evaluate clinical practicability.Results Placenta accreta was confirmed in 56 patients after cesarean section, and in 74 patients without placenta accreta. Based on 1688 omics features included in the image preprocessing, 9 image omics features were selected for the construction of the model after least absolute shrinkage and selection operator (LASSO), Selectbest and REF processing. Five kinds of classifier models in the validation set (LR AUC=0.96, SVM AUC=0.97, RF AUC=0.99, DT AUC=0.87, KNN AUC=0.96) had higher diagnostic efficacy for placenta acta than that of imaging doctors (AUC=0.86). Calibration curves show that the calibration degree of RF models is best in the verification set. When the threshold value of validation set DCA is 0.0-0.6, the clinical net benefit of RF, SVM, KNN and LR models is greater than that of DT models.Conclusions The machine learning model based on MRI T2WI image features can accurately distinguish the presence or absence of placenta accretion, and its diagnostic efficacy is obviously better than that of physicians' visual analysis. In addition, compared with the five models, the RF machine learning model has better performance in the diagnosis of placenta accreta.
[Keywords] placenta accrete spectrum disorders;machine learning;diagnostic efficiency;magnetic resonance imaging

WANG Yingchao1   HUANG Gang2*   BA Zhixia1   XUE Lian3   HUANG Baosheng1   XIA Dongzhou1  

1 Department of Medical Imaging, Zhangye People's Hospital Affiliated to Hexi University, Zhangye 734000, China

2 Department of Radiology, Gansu Province Hospital, Lanzhou 730000, China

3 Department of Obstetrics, Zhangye People's Hospital Affiliated to Hexi University, Zhangye 734000, China

Corresponding author: Huang G, E-mail:

Conflicts of interest   None.

ACKNOWLEDGMENTS University Innovation Fund project of Education Department of Gansu Province (No. 2021B-232); Young Teachers Scientific Research Fund Project of Hexi University (No. QN2020005).
Received  2023-01-03
Accepted  2023-07-21
DOI: 10.12015/issn.1674-8034.2023.08.015
WANG Y C, HUANG G, BA Z X, et al. A preliminary study on diagnostic model of placenta implantation based on magnetic resonance image feature machine learning[J]. Chin J Magn Reson Imaging, 2023, 14(8): 94-99, 107. DOI:10.12015/issn.1674-8034.2023.08.015.

SILVER R M, BRANCH D W. Placenta accreta spectrum[J]. N Engl J Med, 2018, 378(16): 1529-1536. DOI: 10.1056/NEJMcp1709324.
BLOOMFIELD V, ROGERS S, LEYLAND N. Placenta accreta spectrum[J/OL]. CMAJ, 2020, 192(34): E980 [2023-01-01] DOI: 10.1503/cmaj.200304.
MATSUZAKI S, YOSHINO K, ENDO M, et al. Conservative management of placenta percreta[J]. Int J Gynaecol Obstet, 2018, 140(3): 299-306. DOI: 10.1002/ijgo.12411.
SILVER R M, LANDON M B, ROUSE D J, et al. Maternal morbidity associated with multiple repeat cesarean deliveries[J]. Obstet Gynecol, 2006, 107(6): 1226-1232. DOI: 10.1097/01.AOG.0000219750.79480.84.
SCHWICKERT A, VAN BEEKHUIZEN H J, BERTHOLDT C, et al. Association of peripartum management and high maternal blood loss at cesarean delivery for placenta accreta spectrum (PAS): A multinational database study[J]. Acta Obstet Gynecol Scand, 2021, 100(Suppl 1): 29-40. DOI: 10.1111/aogs.14103.
GILLIES R J, KINAHAN P E, HRICAK H. Radiomics: images are more than pictures, they are data[J]. Radiology, 2016, 278(2): 563-577. DOI: 10.1148/radiol.2015151169.
ROMEO V, RICCIARDI C, CUOCOLO R, et al. Machine learning analysis of MRI-derived texture features to predict placenta accreta spectrum in patients with placenta previa[J]. Magn Reson Imaging, 2019, 64: 71-76. DOI: 10.1016/j.mri.2019.05.017.
SUN H Q, QU H B, CHEN L, et al. Identification of suspicious invasive placentation based on clinical MRI data using textural features and automated machine learning[J]. Eur Radiol, 2019, 29(11): 6152-6162. DOI: 10.1007/s00330-019-06372-9.
CHEN E, MAR W A, HOROWITZ J M, et al. Texture analysis of placental MRI: can it aid in the prenatal diagnosis of placenta accreta spectrum?[J]. Abdom Radiol, 2019, 44(9): 3175-3184. DOI: 10.1007/s00261-019-02104-1.
GREENER J G, KANDATHIL S M, MOFFAT L, et al. A guide to machine learning for biologists[J]. Nat Rev Mol Cell Biol, 2022, 23(1): 40-55. DOI: 10.1038/s41580-021-00407-0.
WANG C, ZHANG Y K. Improving scoring-docking-screening powers of protein-ligand scoring functions using random forest[J]. J Comput Chem, 2017, 38(3): 169-177. DOI: 10.1002/jcc.24667.
ZENG W W, WU M M, JIANG R. Prediction of enhancer-promoter interactions via natural language processing[J/OL]. BMC Genomics, 2018, 19(Suppl 2): 84 [2023-01-01]. DOI: 10.1186/s12864-018-4459-6.
BZDOK D, KRZYWINSKI M, ALTMAN N. Machine learning: supervised methods[J]. Nat Methods, 2018, 15(1): 5-6. DOI: 10.1038/nmeth.4551.
JAUNIAUX E, ALFIREVIC Z, BHIDE A G, et al. Placenta praevia and placenta accreta: diagnosis and management: green-top guideline No. 27a[J/OL]. BJOG, 2019, 126(1): e1-e48 [2023-01-01]. DOI: 10.1111/1471-0528.15306.
BOURGIOTI C, ZAFEIROPOULOU K, FOTOPOULOS S, et al. MRI features predictive of invasive placenta with extrauterine spread in high-risk gravid patients: a prospective evaluation[J]. AJR Am J Roentgenol, 2018, 211(3): 701-711. DOI: 10.2214/AJR.17.19303.
FARALLI I, DEL NEGRO V, CHINÈ A, et al. Placenta accreta spectrum (PAS) disorder: ultrasound versus magnetic resonance imaging[J/OL]. Diagnostics, 2022, 12(11): 2769 [2023-01-01]. DOI: 10.3390/diagnostics12112769.
FU F, SHAN Y, YANG G, et al. Deep learning for head and neck CT angiography: Stenosis and plaque classification[J/OL]. Radiology, 2023, 307(3: e220996 [2023-01-01]. DOI: 10.1148/radiol.220996.
FU F, WEI J Y, ZHANG M, et al. Rapid vessel segmentation and reconstruction of head and neck angiograms using 3D convolutional neural network[J/OL]. Nat Commun, 2020, 11(1: 4829 [2023-01-01]. DOI: 10.1038/s41467-020-18606-2.
HUANG F, LAI Q Q, WU H, et al. Application of indirect signs of magnetic resonance imaging (MRI) in prenatal diagnosis of abnormally invasive placenta[J/OL]. Med Sci Monit, 2020, 26: e923272 [2023-01-01]. DOI: 10.12659/MSM.923272.
GUHA A, ANJARI M, COOK G, et al. Radiomic analysis of tumour heterogeneity using MRI in head and neck cancer following chemoradiotherapy: a feasibility study[J/OL]. Front Oncol, 2022, 12: 784693 [2023-01-01]. DOI: 10.3389/fonc.2022.784693.
CONTI A, DUGGENTO A, INDOVINA I, et al. Radiomics in breast cancer classification and prediction[J]. Semin Cancer Biol, 2021, 72: 238-250. DOI: 10.1016/j.semcancer.2020.04.002.
TAHA B, BOLEY D, SUN J, et al. State of radiomics in glioblastoma[J]. Neurosurgery, 2021, 89(2: 177-184. DOI: 10.1093/neuros/nyab124.
KIM M, JUNG S C, PARK J E, et al. Reproducibility of radiomic features in SENSE and compressed SENSE: impact of acceleration factors[J]. Eur Radiol, 2021, 31(9: 6457-6470. DOI: 10.1007/s00330-021-07760-w.
ZHANG M, WONG S W, WRIGHT J N, et al. MRI radiogenomics of pediatric medulloblastoma: a multicenter study[J]. Radiology, 2022, 304(2: 406-416. DOI: 10.1148/radiol.212137.
SCHURINK N W, VAN KRANEN S R, ROBERTI S, et al. Sources of variation in multicenter rectal MRI data and their effect on radiomics feature reproducibility[J]. Eur Radiol, 2022, 32(3: 1506-1516. DOI: 10.1007/s00330-021-08251-8.
OGBONNAYA C N, ZHANG X Y, ALSAEDI B S O, et al. Prediction of clinically significant cancer using radiomics features of pre-biopsy of multiparametric MRI in men suspected of prostate cancer[J/OL]. Cancers, 2021, 13(24: 6199 [2023-01-01]. DOI: 10.3390/cancers13246199.
SANSONE M, GRASSI R, BELFIORE M P, et al. Radiomic features of breast parenchyma: assessing differences between for processing and for presentation digital mammography[J/OL]. Insights Imaging, 2021, 12(1: 147 [2023-01-01]. DOI: 10.1186/s13244-021-01093-4.
KANG Y, MCHUGH M D, CHITTAMS J, et al. Utilizing home healthcare electronic health records for telehomecare patients with heart failure: a decision tree approach to detect associations with rehospitalizations[J]. Comput Inform Nurs, 2016, 34(4: 175-182. DOI: 10.1097/CIN.0000000000000223.
RAJU D, SU X G, PATRICIAN P A, et al. Exploring factors associated with pressure ulcers: a data mining approach[J]. Int J Nurs Stud, 2015, 52(1: 102-111. DOI: 10.1016/j.ijnurstu.2014.08.002.
ALDERDEN J, PEPPER G A, WILSON A, et al. Predicting pressure injury in critical care patients: a machine-learning model[J]. Am J Crit Care, 2018, 27(6: 461-468. DOI: 10.4037/ajcc2018525.
CHEN K, LI R C, DOU Y, et al. Ranking support vector machine with kernel approximation[J/OL]. Comput Intell Neurosci, 2017, 2017: 4629534 [2023-01-01]. DOI: 10.1155/2017/4629534.
NUSINOVICI S, THAM Y C, CHAK YAN M Y, et al. Logistic regression was as good as machine learning for predicting major chronic diseases[J]. J Clin Epidemiol, 2020, 122: 56-69. DOI: 10.1016/j.jclinepi.2020.03.002.
ZHANG S C, LI X L, ZONG M, et al. Efficient kNN classification with different numbers of nearest neighbors[J]. IEEE Trans Neural Netw Learn Syst, 2018, 29(5: 1774-1785. DOI: 10.1109/TNNLS.2017.2673241.
VER HOEF J M, TEMESGEN H. A comparison of the spatial linear model to Nearest Neighbor (k-NN) methods for forestry applications[J/OL]. PLoS One, 2013, 8(3: e59129 [2023-01-01]. DOI: 10.1371/journal.pone.0059129.
DO Q N, LEWIS M A, XI Y, et al. MRI of the placenta accreta spectrum (PAS) disorder: radiomics analysis correlates with surgical and pathological outcome[J]. J Magn Reson Imaging, 2020, 51(3: 936-946. DOI: 10.1002/jmri.26883.

PREV The development and external validation of a model based on MRI quantification, pathology, and blood cell parameters to predict the efficacy of concurrent chemoradiotherapy for stage Ⅱ‍B-‍Ⅲ cervical cancer
NEXT Normalizing radiomics features from multiscale structural MRI of the adolescent brain

Tel & Fax: +8610-67113815    E-mail: