A multi-level similarity measure for the retrieval of the common CT imaging signs of lung diseases

An uncertainty-aware self-training framework with consistency regularization for the multilabel classification of common computed tomography signs in lung nodules

Background

Computed tomography (CT) signs of lung nodules play an important role in indicating lung nodules' malignancy, and accurate automatic classification of these signs can help doctors improve their diagnostic efficiency. However, few relevant studies targeting multilabel classification (MLC) of nodule signs have been conducted. Moreover, difficulty in obtaining labeled data also restricts this avenue of research to a large extent. To address these problems, a multilabel automatic classification system for nodule signs is proposed, which consists of a 3-dimensional (3D) convolutional neural network (CNN) and an efficient new semi-supervised learning (SSL) framework.

Methods

Two datasets were used in our experiments: Lung Nodule Analysis 16 (LUNA16), a public dataset for lung nodule classification, and a private dataset of nodule signs. The private dataset contains 641 nodules, 454 of which were annotated with 6 important signs by radiologists. Our classification system consists of 2 main parts: a 3D CNN model and an SSL method called uncertainty-aware self-training framework with consistency regularization (USC). In the system, supervised training is performed with labeled data, and simultaneously, an uncertainty-and-confidence-based strategy is used to select pseudo-labeled samples for unsupervised training, thus jointly realizing the optimization of the model.

Results

For the MLC of nodule signs, our proposed 3D CNN achieved satisfactory results with a mean average precision (mAP) of 0.870 and a mean area under the curve (AUC) of 0.782. In semi-supervised experiments, compared with supervised learning, our proposed SSL method could increase the mAP by 7.6% (from 0.730 to 0.806) and the mean AUC by 8.1% (from 0.631 to 0.712); it thus efficiently utilized the unlabeled data and achieved superior performance improvement compared to the recently advanced methods.

Conclusions

We realized the optimal MLC of lung nodule signs with our proposed 3D CNN. Our proposed SSL method can also offer an efficient solution for the insufficiency of labeled data that may exist in the MLC tasks of 3D medical images.

Impact of a content-based image retrieval system on the interpretation of chest CTs of patients with diffuse parenchymal lung disease

OBJECTIVES

Content-based image retrieval systems (CBIRS) are a new and potentially impactful tool for radiological reporting, but their clinical evaluation is largely missing. This study aimed at assessing the effect of CBIRS on the interpretation of chest CT scans from patients with suspected diffuse parenchymal lung disease (DPLD).

MATERIALS AND METHODS

A total of 108 retrospectively included chest CT scans with 22 unique, clinically and/or histopathologically verified diagnoses were read by eight radiologists (four residents, four attending, median years reading chest CT scans 2.1± 0.7 and 12 ± 1.8, respectively). The radiologists read and provided the suspected diagnosis at a certified radiological workstation to simulate clinical routine. Half of the readings were done without CBIRS and half with the additional support of the CBIRS. The CBIRS retrieved the most likely of 19 lung-specific patterns from a large database of 6542 thin-section CT scans and provided relevant information (e.g., a list of potential differential diagnoses).

RESULTS

Reading time decreased by 31.3% (p < 0.001) despite the radiologists searching for additional information more frequently when the CBIRS was available (154 [72%] vs. 95 [43%], p < 0.001). There was a trend towards higher overall diagnostic accuracy (42.2% vs 34.7%, p = 0.083) when the CBIRS was available.

CONCLUSION

The use of the CBIRS had a beneficial impact on the reading time of chest CT scans in cases with DPLD. In addition, both resident and attending radiologists were more likely to consult informational resources if they had access to the CBIRS. Further studies are needed to confirm the observed trend towards increased diagnostic accuracy with the use of a CBIRS in practice.

KEY POINTS

• A content-based image retrieval system for supporting the diagnostic process of reading chest CT scans can decrease reading time by 31.3% (p < 0.001). • The decrease in reading time was present despite frequent usage of the content-based image retrieval system. • Additionally, a trend towards higher diagnostic accuracy was observed when using the content-based image retrieval system (42.2% vs 34.7%, p = 0.083).

A self-supervised feature-standardization-block for cross-domain lung disease classification

With the advance of deep learning technology, convolutional neural network (CNN) has been wildly used and achieved the state-of-the-art performances in the area of medical image classification. However, most existing medical image classification methods conduct their experiments on only one public dataset. When applying a well-trained model to a different dataset selected from different sources, the model usually shows large performance degradation and needs to be fine-tuned before it can be applied to the new dataset. The goal of this work is trying to solve the cross-domain image classification problem without using data from target domain. In this work, we designed a self-supervised plug-and-play feature-standardization-block (FSB) which consisting of image normalization (INB), contrast enhancement (CEB) and boundary detection blocks (BDB), to extract cross-domain robust feature maps for deep learning framework, and applied the network for chest x-ray-based lung diseases classification. Three classic deep networks, i.e. VGG, Xception and DenseNet and four chest x-ray lung diseases datasets were employed for evaluating the performance. The experimental result showed that when employing feature-standardization-block, all three networks showed better domain adaption performance. The image normalization, contrast enhancement and boundary detection blocks achieved in average 2%, 2% and 5% accuracy improvement, respectively. By combining all three blocks, feature-standardization-block achieved in average 6% accuracy improvement.

Ensemble learning based automatic detection of tuberculosis in chest X-ray images using hybrid feature descriptors

Tuberculosis (TB) remains one of the major health problems in modern times with a high mortality rate. While efforts are being made to make early diagnosis accessible and more reliable in high burden TB countries, digital chest radiography has become a popular source for this purpose. However, the screening process requires expert radiologists which may be a potential barrier in developing countries. A fully automatic computer-aided diagnosis system can reduce the need of trained personnel for early diagnosis of TB using chest X-ray images. In this paper, we have proposed a novel TB detection technique that combines hand-crafted features with deep features (convolutional neural network-based) through Ensemble Learning. Handcrafted features were extracted via Gabor Filter and deep features were extracted via pre-trained deep learning models. Two publicly available datasets namely (i) Montgomery and (ii) Shenzhen were used to evaluate the proposed system. The proposed methodology was validated with a k-fold cross-validation scheme. The area under receiver operating characteristics curves of 0.99 and 0.97 were achieved for Shenzhen and Montgomery datasets respectively which shows the superiority of the proposed scheme.