Automated interpretation of ventilation-perfusion lung scintigrams for the diagnosis of pulmonary embolism using artificial neural networks

Abstract

The purpose of this study was to develop a completely automated method for the interpretation of ventilation-perfusion (V-P) lung scintigrams used in the diagnosis of pulmonary embolism. An artificial neural network was trained for the diagnosis of pulmonary embolism using 18 automatically obtained features from each set of V-P scintigrams. The techniques used to process the images included their alignment to templates, the construction of quotient images based on the ventilation and perfusion images, and the calculation of measures describing V-P mismatches in the quotient images. The templates represented lungs of normal size and shape without any pathological changes. Images that could not be properly aligned to the templates were detected and excluded automatically. After exclusion of those V-P scintigrams not properly aligned to the templates, 478 V-P scintigrams remained in a training group of consecutive patients with suspected pulmonary embolism, and a further 87 V-P scintigrams formed a separate test group comprising patients who had undergone pulmonary angiography. The performance of the neural network, measured as the area under the receiver operating characteristic curve, was 0.87 (95% confidence limits 0.82-0.92) in the training group and 0.79 (0.69-0.88) in the test group. It is concluded that a completely automated method can be used for the interpretation of V-P scintigrams. The performance of this method is similar to others previously presented, whereby features were extracted manually.

Predicting the Presence of Acute Pulmonary Embolism: A Comparative Analysis of the Artificial Neural Network, Logistic Regression, and Threshold Models

“…Also, I used a single method of encoding the raw data into input variables; a different encoding method potentially could produce different comparative results. However, the area under the ROC curve associated with the neural network in my study, 0.78 (SD, 0.02), is similar to those of previous studies evaluating neural network analysis of ventilation-perfusion scanning data [15][16][17][18][19][20][21] where the mean area under the ROC curve for neural networks was 0.81 (SD, 0.06). This similarity between the performance of the neural networks previously described in the literature and the performance of the neural network in this study exists despite heterogeneity in the neural network structures, encoding of input variables, and patient populations among these studies.…”

“…These studies incorporated comparisons with the diagnostic accuracy of experienced physicians as a performance benchmark. In most cases [15][16][17][18][19][20], the overall performance of the neural network was found to be similar to that of the physicians. The conclusions of these studies may be that the neural network is a successful or promising method of data analysis for the diagnosis of pulmonary embolism; however, whether the neural network is unique in its power to predict the presence of this disease is unknown because none of the studies involved a direct comparison with any other data analysis method.…”

“…Performance of the various artificial neural networks [15][16][17][18][19][20][21] may be similarly accurate in the prediction of the presence of pulmonary embolism, but the models differ according to the way in which their input variables are derived from the imaging data. The first type [19][20][21] relies on input variables that encode observations and interpretations made by humans.…”

“…The first type [19][20][21] relies on input variables that encode observations and interpretations made by humans. The second type [15][16][17][18] relies on computer-generated values derived directly from image processing and is therefore not subject to variations associated with human observation. Human judgment does affect both types of models, however, because humans still determine which input variables are created or chosen when the models are developed.…”

“…The disease continues to be a source of significant morbidity and mortality, and its diagnosis remains challenging [14]. Several artificial neural networks have been developed for the diagnosis of pulmonary embolism from findings observed on ventilation-perfusion scans [15][16][17][18][19][20][21]. These studies incorporated comparisons with the diagnostic accuracy of experienced physicians as a performance benchmark.…”