A promising approach for screening pulmonary hypertension based on frontal chest radiographs using deep learning: A retrospective study

Abstract

Background To date, the missed diagnosis rate of pulmonary hypertension (PH) was high, and there has been limited development of a rapid, simple, and effective way to screen the disease. The purpose of this study is to develop a deep learning approach to achieve rapid detection of possible abnormalities in chest radiographs suggesting PH for screening patients suspected of PH. Methods We retrospectively collected frontal chest radiographs and the pulmonary artery systolic pressure (PASP) value measured by Doppler transthoracic echocardiography from 762 patients (357 healthy controls and 405 with PH) from three institutes in China from January 2013 to May 2019. The wohle sample comprised 762 images (641 for training, 80 for internal test, and 41 for external test). We firstly performed a 8-fold cross-validation on the 641 images selected for training (561 for pre-training, 80 for validation), then decided to tune learning rate to 0.0008 according to the best score on validation data. Finally, we used all the pre-training and validation data (561+80 = 641) to train our models (Resnet50, Xception, and Inception V3), evaluated them on internal and external test dataset to classify the images as having manifestations of PH or healthy according to the area under the receiver operating characteristic curve (AUC/ROC). After that, the three deep learning models were further used for prediction of PASP using regression algorithm. Moreover, we invited an experienced chest radiologist to classify the images in the test dataset as having PH or not, and compared the prediction accuracy performed by deep learing models with that of manual classification.

Assessment of the effect of a comprehensive chest radiograph deep learning model on radiologist reports and patient outcomes: a real-world observational study

“…11 Of note is the rise of convolutional neural networks (CNNs), a type of deep neural network that excels at image feature extraction and classification, and demonstrates strong performance in medical image analysis, leading to the rapid advancement of computer vision in medical imaging. 12 13 CNNs have been used to develop models to successfully detect targeted clinical findings on CXR, including lung cancer, 14 15 pneumonia, 16 17 COVID-19, 18 pneumothorax, [19][20][21][22] pneumoconiosis, 23 cardiomegaly, 24 pulmonary hypertension 25 and tuberculosis. [26][27][28][29][30] These studies highlight the effectiveness of applied machine learning in CXR interpretation, however, most of these deep learning systems are limited in scope to a single finding or a small set of findings, therefore lacking the broad utility that would make them useful in clinical practice.…”

Physical Examination for the Detection of Pulmonary Hypertension: A Systematic Review

“…There has been limited literature on an evidence-based approach for the physical examination in screening or in the early detection of this condition. Screening has been suggested for selected patients with the following conditions: systemic sclerosis, human immunodeficiency virus, congestive heart failure, anorexigenic drug users, and advanced liver disease [6,22,23]. Our study has not identified any single physical examination test to accurately confirm or exclude the diagnosis of PH, and further study will be needed to determine the role of bedside diagnosis with regard to PH.…”

Chest radiographs and machine learning – Past, present and future

“… 15 Recent evidence suggests that machine learning models designed to identify lung cancer on CXR are highly sensitive. 16 Other studies have demonstrated strong performance of narrow models designed to detect pneumonia, 17 pneumothorax, 18 pneumoconiosis, 19 cardiomegaly, 20 pulmonary hypertension 21 and tuberculosis. 22 However, narrow models may be problematic in that they draw attention to the presence or absence of the finding they were trained to detect, which may distract the interpreting clinician from other subtle but salient clinical findings.…”