Prenatal prediction and typing of placental invasion using MRI deep and radiomic features

Abstract

Background To predict placental invasion (PI) and determine the subtype according to the degree of implantation, and to help physicians develop appropriate therapeutic measures, a prenatal prediction and typing of placental invasion method using MRI deep and radiomic features were proposed. Methods The placental tissue of abdominal magnetic resonance (MR) image was segmented to form the regions of interest (ROI) using U-net. The radiomic features were subsequently extracted from ROI. Simultaneously, a deep dynamic convolution neural network (DDCNN) with codec structure was established, which was trained by an autoencoder model to extract the deep features from ROI. Finally, combining the radiomic features and deep features, a classifier based on the multi-layer perceptron model was designed. The classifier was trained to predict prenatal placental invasion as well as determine the invasion subtype. Results The experimental results show that the average accuracy, sensitivity, and specificity of the proposed method are 0.877, 0.857, and 0.954 respectively, and the area under the ROC curve (AUC) is 0.904, which outperforms the traditional radiomic based auxiliary diagnostic methods. Conclusions This work not only labeled the placental tissue of MR image in pregnant women automatically but also realized the objective evaluation of placental invasion, thus providing a new approach for the prenatal diagnosis of placental invasion.

Deep Learning Model Based on Multisequence MRI Images for Assessing Adverse Pregnancy Outcome in Placenta Accreta

“…This paper regarded the whole uterus area (including placenta, myometrium, cervix, and so on) and the bladder area as the ROI area 13 . Figure 2 illustrated the FIESTA sequence image, SSFSE sequence image, and the labeled ROI area on the MRI image.…”

“…This paper regarded the whole uterus area (including placenta, myometrium, cervix, and so on) and the bladder area as the ROI area. 13 Figure 2 illustrated the FIESTA sequence image, SSFSE sequence image, and the labeled ROI area on the MRI image. Three graduate students used the open source labeling software ITK-Snap to draw the ROIs of 21 cases under the supervision of a professional radiologist, 14 they were trained and practiced for several weeks and these manually drawn images are evaluated and corrected by two senior radiologists with more than 10 years' experience.…”

“…There are few PAS disorders studies based on deep learning. 13,33 These studies used deep learning technologies to extract semantic features of placental tissue, and used radiomics analysis to extract radiomic features. Then, a classifier was established to assess PAS based on the extracted semantic features and radiomic features.…”

“…33 Xuan et al also used an autoencoder deep learning model to extract the deep features from ROI and combined the deep features with extracted radiomic features together as the input of a multilayer perceptron classifier model, which was used for assessing the different severity level of placental invasion. 13 However, these studies mainly obtained deep features by training automatic encoder segmentation models, and then needed to train a classifier to assess. Like the radiomic features based machine learning method mentioned earlier, the entire process is divided into two steps, which will lead to the final assessment result is usually locally suboptimal rather than globally suboptimal.…”

A Deep Learning Radiomics Nomogram to Predict Response to Neoadjuvant Chemotherapy for Locally Advanced Cervical Cancer: A Two-Center Study

“…Deep learning (DL) performs well in medical image classification and prediction, extracting and learning deep features directly in a data-driven manner [ 18 , 19 , 20 ], and has shown excellent performance in precision cancer diagnosis, therapeutic decision making and survival analysis [ 21 , 22 , 23 ]. Since DL provides more abstract and difficult-to-mine radiomics features, several researchers have combined handcrafted and DL-based radiomics features to develop prediction models in cancers [ 24 , 25 , 26 ]. Although radiomics is capable of quantifying interpretable tumor image features, DL-based features primarily focus on effective semantic information, making them suitable for large-scale image analysis [ 27 , 28 ].…”