Examining feature extraction and classification modules in machine learning for diagnosis of low-dose computed tomographic screening-detected in vivo lesions.

Purpose: Medical imaging-based machine learning (ML) for computer-aided diagnosis of in vivo lesions consists of two basic components or modules of (i) feature extraction from non-invasively acquired medical images and (ii) feature classification for prediction of malignancy of lesions detected or localized in the medical images. This study investigates their individual performances for diagnosis of low-dose computed tomography (CT) screening-detected lesions of pulmonary nodules and colorectal polyps. Approach: Three feature extraction methods were investigated. One uses the mathematical descriptor of gray-level co-occurrence image texture measure to extract the Haralick image texture features (HFs). One uses the convolutional neural network (CNN) architecture to extract deep learning (DL) image abstractive features (DFs). The third one uses the interactions between lesion tissues and X-ray energy of CT to extract tissue-energy specific characteristic features (TFs). All the above three categories of extracted features were classified by the random forest (RF) classifier with comparison to the DL-CNN method, which reads the images, extracts the DFs, and classifies the DFs in an end-to-end manner. The ML diagnosis of lesions or prediction of lesion malignancy was measured by the area under the receiver operating characteristic curve (AUC). Three lesion image datasets were used. The lesions' tissue pathological reports were used as the learning labels. Results: Experiments on the three datasets produced AUC values of 0.724 to 0.878 for the HFs, 0.652 to 0.965 for the DFs, and 0.985 to 0.996 for the TFs, compared to the DL-CNN of 0.694 to 0.964. These experimental outcomes indicate that the RF classifier performed comparably to the DL-CNN classification module and the extraction of tissue-energy specific characteristic features dramatically improved AUC value. Conclusions: The feature extraction module is more important than the feature classification module. Extraction of tissue-energy specific characteristic features is more important than extraction of image abstractive and characteristic features.

Texture-aware dual domain mapping model for low-dose CT reconstruction.

BACKGROUND: Remarkable progress has been made for low-dose computed tomography (CT) reconstruction tasks by applying deep learning techniques. However, establishing an intrinsic link between deep learning techniques and CT texture preservation is still one of the significant challenges for researchers to further improve the effect of low-dose CT (LDCT) reconstruction. PURPOSE: Most of the existing deep learning-based LDCT reconstruction methods are derived from popular frameworks, and most models focus on the image domain. Even few existing methods start with dual domains (sinogram and image) by considering the processing of the data itself, the final performances are limited due to the lack of perception of textures. With this in mind, we propose a method for texture perception on dual domains, so that the reconstruction process can be uniformly driven by visual effects. METHODS: The proposed method involves the processing of two domains: the sinogram domain and the image domain. For the sinogram domain, we have designed a novel dilated residual network (S-DRN) which aims to increase the receptive field to obtain multiscale information. For the image domain, we propose a self-attention (SA) residual encoder & decoder network (SRED-Net) as the denoising network for obtaining much acceptable edges and textures. In addition, the composite loss function composed of the feature loss constructed by the proposed boundary and texture feature-aware network (BTFAN) and the mean square error (MSE) can obtain a higher image quality while retaining more details and fewer artifacts, thereby obtaining better visual image quality. RESULTS: The proposed method was validated using both the American association of physicists in medicine (AAPM)-Mayo clinic LDCT data sets and a real clinic data. Experimental results demonstrated that the new method has achieved the state-of-the-art performance on objective indicators and visual metrics in terms of denoising and texture restoration. CONCLUSIONS: Compared with single-domain or existing dual-domain processing strategies, the proposed texture-aware dual domain mapping network (TADDM-Net) can much better improve the visual effect of reconstructed CT images. Meantime, we also provide much intuitive evidence in terms of model interpretability.

Fully-automatic segmentation of coronary artery using growing algorithm.

Currently, cardiac computed tomography angiography (CTA) is widely applied to coronary artery disease diagnosis. Automatic segmentation of coronary artery has played an important role in coronary artery disease diagnosis. In this study, we propose and test a fully automatic coronary artery segmentation method that does not require any human-computer interaction. The proposed method uses a growing strategy and contains three main parts namely, (1) the initial seed detection that automatically detects the root points of the left and right coronary arteries where the ascending aorta meets the coronary arteries, (2) the growing strategy that searches for the neighborhood blocks to decide the existence of coronary arteries with an improved convolutional neural network, and (3) the iterative termination condition that decides whether the growing iteration finishes. The proposed framework is validated using a dataset containing 32 cardiac CTA volumes from different patients for training and testing. Experimental results show that the proposed method obtained a Dice loss ranged from 0.70 to 0.83, which indicates that the new method outperforms the traditional methods such as level set.

An adaptive paradigm for computer-aided detection of colonic polyps.

Most previous efforts in developing computer-aided detection (CADe) of colonic polyps apply similar measures or parameters to detect polyps regardless of their locations under an implicit assumption that all the polyps reside in a similar local environment, e.g. on a relatively flat colon wall. In reality, this implicit assumption is frequently invalid, because the haustral folds can have a very different local environment from that of the relatively flat colon wall. We conjecture that this assumption may be a major cause of missing the detection of polyps, especially small polyps (<10 mm linear size) located on the haustral folds. In this paper, we take the concept of adaptiveness and present an adaptive paradigm for CADe of colonic polyps. Firstly, we decompose the complicated colon structure into two simplified sub-structures, each of which has similar properties, of (1) relatively flat colon wall and (2) ridge-shaped haustral folds. Then we develop local environment descriptions to adaptively reflect each of these two simplified sub-structures. To show the impact of the adaptiveness of the local environment descriptions upon the polyp detection task, we focus on the local geometrical measures of the volume data for both the detection of initial polyp candidates (IPCs) and the reduction of false positives (FPs) in the IPC pool. The experimental outcome using the local geometrical measures is very impressive such that not only the previously-missed small polyps on the folds are detected, but also the previously miss-removed small polyps on the folds during FP reduction are retained. It is expected that this adaptive paradigm will have a great impact on detecting the small polyps, measuring their volumes and volume changes over time, and optimizing their management plan.