Automatic Categorization and Scoring of Solid, Part-Solid and Non-Solid Pulmonary Nodules in CT Images with Convolutional Neural Network

Abstract

We present a computer-aided diagnosis system (CADx) for the automatic categorization of solid, part-solid and non-solid nodules in pulmonary computerized tomography images using a Convolutional Neural Network (CNN). Provided with only a two-dimensional region of interest (ROI) surrounding each nodule, our CNN automatically reasons from image context to discover informative computational features. As a result, no image segmentation processing is needed for further analysis of nodule attenuation, allowing our system to avoid potential errors caused by inaccurate image processing. We implemented two computerized texture analysis schemes, classification and regression, to automatically categorize solid, part-solid and non-solid nodules in CT scans, with hierarchical features in each case learned directly by the CNN model. To show the effectiveness of our CNN-based CADx, an established method based on histogram analysis (HIST) was implemented for comparison. The experimental results show significant performance improvement by the CNN model over HIST in both classification and regression tasks, yielding nodule classification and rating performance concordant with those of practicing radiologists. Adoption of CNN-based CADx systems may reduce the inter-observer variation among screening radiologists and provide a quantitative reference for further nodule analysis.

Highly accurate model for prediction of lung nodule malignancy with CT scans

“…For example, with a large number of images, the learning model for face recognition of Facebook was trained on four million images and it is said that the model reached an accuracy level even higher than the FBI current system for face recognition. Other groups, including 20 – 23 have trained CNN models with fewer layers; this would help alleviate the training difficulty by reducing the total parameter count. As pointed out in 22 , it is possible that deeper CNNs may not offer much improvement over more shallow models.…”

“…We chose a tradeoff to reduce a potential information-leakage bias, but acknowledge that this approach is susceptible to other biases. For comparison 20 – 23 , performed experiments on the LIDC/IDRI dataset with CNN models and chose to use cross-validation. (iii) For training and validating we use the LIDC/IDRI dataset.…”

“…Deep learning convolutional neural network (CNN) based approaches are very promising with the availability of CT scans from large cohorts. Many recent efforts are devoted to nodule classification approaches using convolutional neural networks 16 – 23 . The LIDC/IDRI cohort has been used by the authors of 16 – 23 to train and test their models for classifying lung nodules.…”

Evolving the pulmonary nodules diagnosis from classical approaches to deep learning-aided decision support: three decades’ development course and future prospect

“…Tu et al (Tu et al, 2017) implemented classification to automatically categorize solid, part-solid and non-solid pulmonary nodules in CT scans by directly using the hierarchical features learned from the CNN. They conducted a comprehensive performance comparison between the CNN-based categorization method and the histogram-based approach (HIST).…”

Deep learning aided decision support for pulmonary nodules diagnosing: a review

“…Tu et al (14) trained a two-dimensional CNN for categorization of solid, part-solid and non-solid pulmonary nodules in computed tomography (CT) images. The CNN was structured with two convolutional layers (5×5 kernel size with stride 2), each followed by a max-pooling layer (2×2 kernel size with stride 2).…”