Applicability of artificial intelligence-based computer-aided detection (AI-CAD) for pulmonary tuberculosis to community-based active case finding.

BACKGROUND: Artificial intelligence-based computer-aided detection (AI-CAD) for tuberculosis (TB) has become commercially available and several studies have been conducted to evaluate the performance of AI-CAD for pulmonary tuberculosis (TB) in clinical settings. However, little is known about its applicability to community-based active case-finding (ACF) for TB. METHODS: We analysed an anonymized data set obtained from a community-based ACF in Cambodia, targeting persons aged 55 years or over, persons with any TB symptoms, such as chronic cough, and persons at risk of TB, including household contacts. All of the participants in the ACF were screened by chest radiography (CXR) by Cambodian doctors, followed by Xpert test when they were eligible for sputum examination. Interpretation by an experienced chest physician and abnormality scoring by a newly developed AI-CAD were retrospectively conducted for the CXR images. With a reference of Xpert-positive TB or human interpretations, receiver operating characteristic (ROC) curves were drawn to evaluate the AI-CAD performance by area under the ROC curve (AUROC). In addition, its applicability to community-based ACFs in Cambodia was examined. RESULTS: TB scores of the AI-CAD were significantly associated with the CXR classifications as indicated by the severity of TB disease, and its AUROC as the bacteriological reference was 0.86 (95% confidence interval 0.83-0.89). Using a threshold for triage purposes, the human reading and bacteriological examination needed fell to 21% and 15%, respectively, detecting 95% of Xpert-positive TB in ACF. For screening purposes, we could detect 98% of Xpert-positive TB cases. CONCLUSIONS: AI-CAD is applicable to community-based ACF in high TB burden settings, where experienced human readers for CXR images are scarce. The use of AI-CAD in developing countries has the potential to expand CXR screening in community-based ACFs, with a substantial decrease in the workload on human readers and laboratory labour. Further studies are needed to generalize the results to other countries by increasing the sample size and comparing the AI-CAD performance with that of more human readers.

Decomposition of musculoskeletal structures from radiographs using an improved CycleGAN framework.

This paper presents methods of decomposition of musculoskeletal structures from radiographs into multiple individual muscle and bone structures. While existing solutions require dual-energy scan for the training dataset and are mainly applied to structures with high-intensity contrast, such as bones, we focused on multiple superimposed muscles with subtle contrast in addition to bones. The decomposition problem is formulated as an image translation problem between (1) a real X-ray image and (2) multiple digitally reconstructed radiographs, each of which contains a single muscle or bone structure, and solved using unpaired training based on the CycleGAN framework. The training dataset was created via automatic computed tomography (CT) segmentation of muscle/bone regions and virtually projecting them with geometric parameters similar to the real X-ray images. Two additional features were incorporated into the CycleGAN framework to achieve a high-resolution and accurate decomposition: hierarchical learning and reconstruction loss with the gradient correlation similarity metric. Furthermore, we introduced a new diagnostic metric for muscle asymmetry directly measured from a plain X-ray image to validate the proposed method. Our simulation and real-image experiments using real X-ray and CT images of 475 patients with hip diseases suggested that each additional feature significantly enhanced the decomposition accuracy. The experiments also evaluated the accuracy of muscle volume ratio measurement, which suggested a potential application to muscle asymmetry assessment from an X-ray image for diagnostic and therapeutic assistance. The improved CycleGAN framework can be applied for investigating the decomposition of musculoskeletal structures from single radiographs.

Translation of Cellular Protein Localization Using Convolutional Networks.

Protein localization in cells has been analyzed by fluorescent labeling using indirect immunofluorescence and fluorescent protein tagging. However, the relationships between the localization of different proteins had not been analyzed using artificial intelligence. Here, we applied convolutional networks for the prediction of localization of the cytoskeletal proteins from the localization of the other proteins. Lamellipodia are one of the actin-dependent subcellular structures involved in cell migration and are mainly generated by the Wiskott-Aldrich syndrome protein (WASP)-family verprolin homologous protein 2 (WAVE2) and the membrane remodeling I-BAR domain protein IRSp53. Focal adhesion is another actin-based structure that contains vinculin protein and promotes lamellipodia formation and cell migration. In contrast, microtubules are not directly related to actin filaments. The convolutional network was trained using images of actin filaments paired with WAVE2, IRSp53, vinculin, and microtubules. The generated images of WAVE2, IRSp53, and vinculin were highly similar to their real images. In contrast, the microtubule images generated from actin filament images were inferior without the generation of filamentous structures, suggesting that microscopic images of actin filaments provide more information about actin-related protein localization. Collectively, this study suggests that image translation by the convolutional network can predict the localization of functionally related proteins, and the convolutional network might be used to describe the relationships between the proteins by their localization.

2D-3D reconstruction of distal forearm bone from actual X-ray images of the wrist using convolutional neural networks.

The purpose of the study was to develop a deep learning network for estimating and constructing highly accurate 3D bone models directly from actual X-ray images and to verify its accuracy. The data used were 173 computed tomography (CT) images and 105 actual X-ray images of a healthy wrist joint. To compensate for the small size of the dataset, digitally reconstructed radiography (DRR) images generated from CT were used as training data instead of actual X-ray images. The DRR-like images were generated from actual X-ray images in the test and adapted to the network, and high-accuracy estimation of a 3D bone model from a small data set was possible. The 3D shape of the radius and ulna were estimated from actual X-ray images with accuracies of 1.05 ± 0.36 and 1.45 ± 0.41 mm, respectively.

Automated Surgical Instrument Detection from Laparoscopic Gastrectomy Video Images Using an Open Source Convolutional Neural Network Platform.

BACKGROUND: The common use of laparoscopic intervention produces impressive amounts of video data that are difficult to review for surgeons wishing to evaluate and improve their skills. Therefore, a need exists for the development of computer-based analysis of laparoscopic video to accelerate surgical training and assessment. We developed a surgical instrument detection system for video recordings of laparoscopic gastrectomy procedures. This system, the use of which might increase the efficiency of the video reviewing process, is based on the open source neural network platform, YOLOv3. STUDY DESIGN: A total of 10,716 images extracted from 52 laparoscopic gastrectomy videos were included in the training and validation data sets. We performed 200,000 iterations of training. Video recordings of 10 laparoscopic gastrectomies, independent of the training and validation data set, were analyzed by our system, and heat maps visualizing trends of surgical instrument usage were drawn. Three skilled surgeons evaluated whether each heat map represented the features of the corresponding operation. RESULTS: After training, the testing data set precision and sensitivity (recall) was 0.87 and 0.83, respectively. The heat maps perfectly represented the devices used during each operation. Without reviewing the video recordings, the surgeons accurately recognized the type of anastomosis, time taken to initiate duodenal and gastric dissection, and whether any irregular procedure was performed, from the heat maps (correct answer rates ≥ 90%). CONCLUSIONS: A new automated system to detect manipulation of surgical instruments in video recordings of laparoscopic gastrectomies based on the open source neural network platform, YOLOv3, was developed and validated successfully.

Fully automatic estimation of pelvic sagittal inclination from anterior-posterior radiography image using deep learning framework.

BACKGROUND AND OBJECTIVE: Malposition of the acetabular component causes dislocation and prosthetic impingement after Total Hip Arthroplasty (THA), which significantly affects the postoperative quality of life and implant longevity. The position of the acetabular component is determined by the Pelvic Sagittal Inclination (PSI), which not only varies among different people but also changes in different positions. It is important to recognize individual dynamic changes of the PSI for patient-specific planning of the THA. Previously PSI was estimated by registering the CT and radiography images. In this study, we introduce a new method for accurate estimation of functional PSI without requiring CT image in order to lower radiation exposure of the patient which opens up the possibility of increasing its application in a larger number of hospitals where CT is not acquired as a routine protocol. METHODS: The proposed method consists of two main steps: First, the Mask R-CNN framework was employed to segment the pelvic shape from the background in the radiography images. Then, following the segmentation network, another convolutional network regressed the PSI angle. We employed a transfer learning paradigm where the network weights were initialized by non-medical images followed by fine-tuning using radiography images. Furthermore, in the training process, augmented data was generated to improve the performance of both networks. We analyzed the role of segmentation network in our system and investigated the Mask R-CNN performance in comparison with the U-Net, which is commonly used for the medical image segmentation. RESULTS: In this study, the Mask R-CNN utilizing multi-task learning, transfer learning, and data augmentation techniques achieve 0.960 ± 0.008 DICE coefficient, which significantly outperforms the U-Net. The cascaded system is capable of estimating the PSI with 4.04° ± 3.39° error for the radiography images. CONCLUSIONS: The proposed framework suggests a fully automatic and robust estimation of the PSI using only an anterior-posterior radiography image.

Automated Muscle Segmentation from Clinical CT Using Bayesian U-Net for Personalized Musculoskeletal Modeling.

We propose a method for automatic segmentation of individual muscles from a clinical CT. The method uses Bayesian convolutional neural networks with the U-Net architecture, using Monte Carlo dropout that infers an uncertainty metric in addition to the segmentation label. We evaluated the performance of the proposed method using two data sets: 20 fully annotated CTs of the hip and thigh regions and 18 partially annotated CTs that are publicly available from The Cancer Imaging Archive (TCIA) database. The experiments showed a Dice coefficient (DC) of 0.891±0.016 (mean±std) and an average symmetric surface distance (ASD) of 0.994±0.230 mm over 19 muscles in the set of 20 CTs. These results were statistically significant improvements compared to the state-of-the-art hierarchical multi-atlas method which resulted in 0.845 ± 0.031 DC and 1.556 ± 0.444 mm ASD. We evaluated validity of the uncertainty metric in the multi-class organ segmentation problem and demonstrated a correlation between the pixels with high uncertainty and the segmentation failure. One application of the uncertainty metric in active-learning is demonstrated, and the proposed query pixel selection method considerably reduced the manual annotation cost for expanding the training data set. The proposed method allows an accurate patient-specific analysis of individual muscle shapes in a clinical routine. This would open up various applications including personalization of biomechanical simulation and quantitative evaluation of muscle atrophy.

Analysis of forearm rotational motion using biplane fluoroscopic intensity-based 2D-3D matching.

Measuring three-dimensional (3D) forearm rotational motion is difficult. We aimed to develop and validate a new method for analyzing 3D forearm rotational motion. We proposed biplane fluoroscopic intensity-based 2D-3D matching, which employs automatic registration processing using the evolutionary optimization strategy. Biplane fluoroscopy was conducted for forearm rotation at 12.5 frames per second along with computed tomography (CT) at one static position. An arm phantom was embedded with eight stainless steel spheres (diameter, 1.5 mm), and forearm rotational motion measurements using the proposed method were compared with those using radiostereometric analysis, which is considered the ground truth. As for the time resolution analysis, we measured radiohumeral joint motion in a patient with posterolateral rotatory instability and compared the 2D-3D matching method with the simulated multiple CT method, which uses CTs at multiple positions and interpolates between the positions. Rotation errors of the radius and ulna between these two methods were 0.31 ±â€¯0.35° and 0.32 ±â€¯0.33°, respectively, translation errors were 0.43 ±â€¯0.35 mm and 0.29 ±â€¯0.25 mm, respectively. Although the 2D-3D method could detect joint dislocation, the multiple CT method could not detect quick motion during joint dislocation. The proposed method enabled high temporal- and spatial-resolution motion analyses with low radiation exposure. Moreover, it enabled the detection of a sudden motion, such as joint dislocation, and may contribute to 3D motion analysis, including joint dislocation, which currently cannot be analyzed using conventional methods.

Recovery of 3D rib motion from dynamic chest radiography and CT data using local contrast normalization and articular motion model.

Dynamic chest radiography (2D x-ray video) is a low-dose and cost-effective functional imaging method with high temporal resolution. While the analysis of rib-cage motion has been shown to be effective for evaluating respiratory function, it has been limited to 2D. We aim at 3D rib-motion analysis for high temporal resolution while keeping the radiation dose at a level comparable to conventional examination. To achieve this, we developed a method for automatically recovering 3D rib motion based on 2D-3D registration of x-ray video and single-time-phase computed tomography. We introduce the following two novel components into the conventional intensity-based 2D-3D registration pipeline: (1) a rib-motion model based on a uniaxial joint to constrain the search space and (2) local contrast normalization (LCN) as a pre-process of x-ray video to improve the cost function of the optimization parameters, which is often called the landscape. The effects of each component on the registration results were quantitatively evaluated through experiments using simulated images and real patients' x-ray videos obtained in a clinical setting. The rotation-angle error of the rib and the mean projection contour distance (mPCD) were used as the error metrics. The simulation experiments indicate that the proposed uniaxial joint model improved registration accuracy. By searching the rotation axis along with the rotation angle of the ribs, the rotation-angle error and mPCD significantly decreased from 2.246 ±â€¯1.839° and 1.148 ±â€¯0.743 mm to 1.495 ±â€¯0.993° and 0.742 ±â€¯0.281 mm, compared to simply applying De Troyer's model. The real-image experiments with eight patients demonstrated that LCN improved the cost function space; thus, robustness in optimization resulting in an average mPCD of 1.255 ±â€¯0.615 mm. We demonstrated that an anatomical-knowledge based constraint and an intensity normalization, LCN, significantly improved robustness and accuracy in rib-motion reconstruction using chest x-ray video.

Quantitative comparison of operative skill using 2- and 3-dimensional monitors during laparoscopic phantom tasks.

BACKGROUND: The recent development of stereoscopic images using 3-dimensional monitors is expected to improve techniques for laparoscopic operation. Several studies have reported technical advantages in using 3-dimensional monitors with regard to operative accuracy and working speed, but there are few reports that analyze forceps motions by 3-dimensional optical tracking systems during standardized laparoscopic phantom tasks. We attempted to develop a 3-dimensional motion analysis system for assessing laparoscopic tasks and to clarify the efficacy of using stereoscopic images from a 3-dimensional monitor to track forceps movement during laparoscopy. METHODS: Twenty surgeons performed 3 tasks (Task 1: a simple operation by the dominant hand, Task 2: a simple operation using both hands, Task 3: a complicated operation using both hands) under 2-dimensional and 3-dimensional systems. We tracked and recorded the motion of forceps tips with an optical marker captured by a 3-dimensional position tracker. We analyzed factors such as forceps path lengths, operation times, and technical errors for each task and compared the results of 2-dimensional and 3-dimensional monitors. RESULTS: Mean operation times and technical errors were improved significantly for all tasks performed under the 3-dimensional system compared with the 2-dimensional system; in addition, mean path lengths for the forceps tips were shorter for all tasks performed under the 3-dimensional system. CONCLUSION: We found that stereoscopic images using a 3-dimensional monitor improved operative techniques with regard to increased accuracy and shorter path lengths for forceps movement, which resulted in a shorter operation time for basic phantom laparoscopic tasks.