Pulmonary nodule detection using hybrid two-stage 3D CNNs

Data augmentation for medical imaging: A systematic literature review

Recent advances in Deep Learning have largely benefited from larger and more diverse training sets. However, collecting large datasets for medical imaging is still a challenge due to privacy concerns and labeling costs. Data augmentation makes it possible to greatly expand the amount and variety of data available for training without actually collecting new samples. Data augmentation techniques range from simple yet surprisingly effective transformations such as cropping, padding, and flipping, to complex generative models. Depending on the nature of the input and the visual task, different data augmentation strategies are likely to perform differently. For this reason, it is conceivable that medical imaging requires specific augmentation strategies that generate plausible data samples and enable effective regularization of deep neural networks. Data augmentation can also be used to augment specific classes that are underrepresented in the training set, e.g., to generate artificial lesions. The goal of this systematic literature review is to investigate which data augmentation strategies are used in the medical domain and how they affect the performance of clinical tasks such as classification, segmentation, and lesion detection. To this end, a comprehensive analysis of more than 300 articles published in recent years (2018-2022) was conducted. The results highlight the effectiveness of data augmentation across organs, modalities, tasks, and dataset sizes, and suggest potential avenues for future research.

PUNDIT: Pulmonary nodule detection with image category transformation

BACKGROUND

Convolutional neural networks (CNNs) have achieved great success in pulmonary nodules detection, which plays an important role in lung cancer screening.

PURPOSE

In this paper, we proposed a novel strategy for pulmonary nodule detection by learning it from a harder task, which was to transform nodule images into normal images. We named this strategy as pulmonary nodule detection with image category transformation (PUNDIT).

METHODS

There were two steps for nodules detection, nodule candidate detection and false positive (FP) reduction. In nodule candidate detection step, a segmentation-based framework was built for detection. We designed an image category transformation (ICT) task to translate nodule images into pixel-to-pixel normal images and share the information of detection and transformation tasks by multitask learning. As for references of transformation tasks, we proposed background consistency losses into standard cycle-consistent adversarial networks, which can solve the problem of background uncontrolled changing. A three-dimensional network was used in FP reduction step.

RESULTS

PUNDIT was evaluated in two datasets, cancer screening dataset (CSD) with 1186 nodules for cross-validation and (CTD) with 3668 nodules for external test. Results were mainly evaluated by competition performance metric (CPM), the average sensitivity at seven predefined FP rates. The CPM was improved from 0.906 to 0.931 in CSD, and from 0.835 to 0.848 in CTD.

CONCLUSIONS

Experimental results showed that PUNDIT can improve the performance of pulmonary nodules detection effectively.

Practice toward standardized performance testing of computer-aided detection algorithms for pulmonary nodule

This study aimed at implementing practice to build a standardized protocol to test the performance of computer-aided detection (CAD) algorithms for pulmonary nodules. A test dataset was established according to a standardized procedure, including data collection, curation and annotation. Six types of pulmonary nodules were manually annotated as reference standard. Three specific rules to match algorithm output with reference standard were applied and compared. These rules included: (1) "center hit" [whether the center of algorithm highlighted region of interest (ROI) hit the ROI of reference standard]; (2) "center distance" (whether the distance between algorithm highlighted ROI center and reference standard center was below a certain threshold); (3) "area overlap" (whether the overlap between algorithm highlighted ROI and reference standard was above a certain threshold). Performance metrics were calculated and the results were compared among ten algorithms under test (AUTs). The test set currently consisted of CT sequences from 593 patients. Under "center hit" rule, the average recall rate, average precision, and average F₁ score of ten algorithms under test were 54.68, 38.19, and 42.39%, respectively. Correspondingly, the results under "center distance" rule were 55.43, 38.69, and 42.96%, and the results under "area overlap" rule were 40.35, 27.75, and 31.13%. Among the six types of pulmonary nodules, the AUTs showed the highest miss rate for pure ground-glass nodules, with an average of 59.32%, followed by pleural nodules and solid nodules, with an average of 49.80 and 42.21%, respectively. The algorithm testing results changed along with specific matching methods adopted in the testing process. The AUTs showed uneven performance on different types of pulmonary nodules. This centralized testing protocol supports the comparison between algorithms with similar intended use, and helps evaluate algorithm performance.

Fracture R-CNN: An anchor-efficient anti-interference framework for skull fracture detection in CT images

BACKGROUND

Skull fracture, as a common traumatic brain injury, can lead to multiple complications including bleeding, leaking of cerebrospinal fluid, infection, and seizures. Automatic skull fracture detection (SFD) is of great importance, especially in emergency medicine.

PURPOSE

Existing algorithms for SFD, developed based on hand-crafted features, suffer from low detection accuracy due to poor generalizability to unseen samples. Deploying deep detectors designed for natural images like Faster Region-based Convolutional Neural Network (R-CNN) for SFD can be helpful but are of high redundancy and with nonnegligible false detections due to the cranial suture and skull base interference. Therefore, we, for the first time, propose an anchor-efficient anti-interference deep learning framework named Fracture R-CNN for accurate SFD with low computational cost.

METHODS

The proposed Fracture R-CNN is developed by incorporating the prior knowledge utilized in clinical diagnosis into the original Faster R-CNN. Specifically, based on the distributions of skull fractures, we first propose an adaptive anchoring region proposal network (AA-RPN) to generate proposals for diverse-scale fractures with low computational complexity. Then, based on the prior knowledge that cranial sutures exist in the junctions of bones and usually contain sclerotic margins, we design an anti-interference head (A-Head) network to eliminate the cranial suture interference for better SFD detection. In addition, to further enhance the anti-interference ability of the proposed A-Head, a difficulty-balanced weighted loss function is proposed to emphasize more on distinguishing the interference areas from the skull base and the cranial sutures during training.

RESULTS

Experimental results demonstrate that the proposed Fracture R-CNN outperforms the current state-of-the-art (SOTA) deep detectors for SFD with a higher recall and fewer false detections. Compared to Faster R-CNN, the proposed Fracture R-CNN improves the average precision (AP) by 11.74% and the free-response receiver operating characteristic (FROC) score by 11.08%. Through validating on various backbones, we further demonstrate the architecture independence of Fracture R-CNN, making it extendable to other detection applications.

CONCLUSIONS

As the customized deep learning-based framework for SFD, Fracture R-CNN can effectively overcome the unique challenges in SFD with less computational cost, leading to a better detection performance compared to the SOTA deep detectors. Moreover, we believe the prior knowledge explored for Fracture R-CNN would shed new light on future deep learning approaches for SFD.

Towards Machine Learning-Aided Lung Cancer Clinical Routines: Approaches and Open Challenges

Advancements in the development of computer-aided decision (CAD) systems for clinical routines provide unquestionable benefits in connecting human medical expertise with machine intelligence, to achieve better quality healthcare. Considering the large number of incidences and mortality numbers associated with lung cancer, there is a need for the most accurate clinical procedures; thus, the possibility of using artificial intelligence (AI) tools for decision support is becoming a closer reality. At any stage of the lung cancer clinical pathway, specific obstacles are identified and "motivate" the application of innovative AI solutions. This work provides a comprehensive review of the most recent research dedicated toward the development of CAD tools using computed tomography images for lung cancer-related tasks. We discuss the major challenges and provide critical perspectives on future directions. Although we focus on lung cancer in this review, we also provide a more clear definition of the path used to integrate AI in healthcare, emphasizing fundamental research points that are crucial for overcoming current barriers.

3D gray density coding feature for benign-malignant pulmonary nodule classification on chest CT

PURPOSE

Early detection is significant to reduce lung cancer-related death. Computer-aided detection system (CADs) can help radiologists to make an early diagnosis. In this paper, we propose a novel 3D gray density coding feature (3D GDC) and fuse it with extracted geometric features. The fusion feature and random forest are used for benign-malignant pulmonary nodule classification on Chest CT.

METHODS

First, a dictionary model is created to acquire codebook. It is used to obtain feature descriptors and includes 3D block database (BD) and distance matrix clustering centers. 3D BD is balanced and randomly selecting from benign and malignant pulmonary nodules of training data. Clustering centers is got by clustering the distance matrix, which is the distance between every two blocks in 3D BD. Then, feature descriptor is obtained by coding the pulmonary nodule with codebook, and 3D GDC feature is the result of histogram statistics on feature descriptor. Second, geometric features are extracted for fusion feature. Finally, random forest is performed for benign-malignant pulmonary nodule classification with fusion feature of the 3D gray density coding feature and the geometric features.

RESULTS

We verify the effectiveness of our method on the public LIDC-IDRI dataset and the private ZSHD dataset. For LIDC-IDRI dataset, compared with other state-of-the-art methods, we achieve more satisfactory results with 93.17 ± 1.94% for accuracy and 97.53 ± 1.62% for AUC. As for private ZSHD dataset, it contains a total of 238 lung nodules from 203 patients. The accuracy and AUC achieved by our method are 90.0% and 93.15%.

CONCLUSIONS

The results show that our method can provide doctors with more accurate results of benign-malignant pulmonary nodule classification for auxiliary diagnosis, and our method is more interpretable than 3D CNN methods, which can provide doctors with more auxiliary information.

Pulmonary Nodule Detection Based on Three-Dimensional Multiscale Convolutional Neural Network with Channel and Spatial Attention

Early screening for pulmonary nodules is currently an important means for reducing lung cancer mortality. In recent years, three-dimensional convolutional neural networks have achieved great success in the field of pulmonary nodule detection. This paper proposes a pulmonary nodule detection method based on a threedimensional multiscale convolutional neural network with channel and spatial attention. First, a multiscale module is designed to extract the image features at different scales. Second, a channel and spatial attention module is designed to mine the correlation information between features from the perspective of space and channel. Then the extracted features are sent to a pyramid-like fusion mechanism, so that the features contain both deep semantic information and shallow position information, which is conducive to object positioning and bounding box regression. In general, the experiments on the LUng Nodule Analysis 2016 (LUNA16) dataset show that the average free-response receiver operating characteristic (FROC) score is 0.846. Compared with other current advanced methods, the method is competitive and effective.