Interpretation of Mammogram and Chest X-Ray Reports Using Deep Neural Networks - Preliminary Results

Abstract

Radiology reports are an important means of communication between radiologists and other physicians. These reports express a radiologist's interpretation of a medical imaging examination and are critical in establishing a diagnosis and formulating a treatment plan. In this paper, we propose a Bidirectional convolutional neural network (Bi-CNN) model for the interpretation and classification of mammograms based on breast density and chest radiographic radiology reports based on the basis of chest pathology. The proposed approach is a part of an auditing system that evaluates radiology reports to decrease incorrect diagnoses. Our study revealed that the proposed Bi-CNN outperforms convolutional neural network, random forest and support vector machine methods.

Recent Advances in Recurrent Neural Networks

“…The same sequence of feature vectors can also be represented as a feature matrix (i.e., an image) to be fed as input to a CNN. CNNs are used in [95] to classify radiology reports. The proposed model is particularly developed for chest pathology and mammogram reports.…”

Justifying diagnosis decisions by deep neural networks

“…Secondly, the ability of GANs to produce ex-amples that don't occur in the training data is essential to the visual justification that will be explained in section 3. 4.…”

“…While there have been many recent attempts to classify X-Rays according to diagnosis with (deep) neural networks ( [2], [3], [4]) current solutions often have limitations or are difficult to interpret. The amount of labeled data in such applications is frequently limited and multiple labels per sample are often present which complicates the diagnosis task.…”