Investigation of Appropriate Scaling of Networks and Images for Convolutional Neural Network-Based Nerve Detection in Ultrasound-Guided Nerve Blocks.

Ultrasound imaging is an essential tool in anesthesiology, particularly for ultrasound-guided peripheral nerve blocks (US-PNBs). However, challenges such as speckle noise, acoustic shadows, and variability in nerve appearance complicate the accurate localization of nerve tissues. To address this issue, this study introduces a deep convolutional neural network (DCNN), specifically Scaled-YOLOv4, and investigates an appropriate network model and input image scaling for nerve detection on ultrasound images. Utilizing two datasets, a public dataset and an original dataset, we evaluated the effects of model scale and input image size on detection performance. Our findings reveal that smaller input images and larger model scales significantly improve detection accuracy. The optimal configuration of model size and input image size not only achieved high detection accuracy but also demonstrated real-time processing capabilities.

Detecting the location of lung cancer on thoracoscopic images using deep convolutional neural networks.

OBJECTIVES: The prevalence of minimally invasive surgeries has increased the need for tumor detection using thoracoscopic images during lung cancer surgery. We conducted this study to analyze the efficacy of a deep convolutional neural network (DCNN) for tumor detection using recorded thoracoscopic images of pulmonary surfaces. MATERIALS AND METHODS: We collected 644 intraoperative thoracoscopic images of changes in pulmonary appearance from 427 patients with lung cancer between 2012 and 2021. The lesion areas on the thoracoscopic images were detected by bounding boxes using an advanced version of YOLO, a well-known DCNN for object detection. The DCNN model was trained and evaluated by a 15-fold cross-validation scheme. Each predicted bounding box was considered successful detection when it overlapped more than 50% of the lesion areas annotated by board-certified surgeons. RESULTS AND CONCLUSIONS: Precision, recall, and F1-measured values of 91.9%, 90.5%, and 91.1%, respectively, were obtained. The presence of lymphatic vessel invasion was associated with successful detection (p = 0.045). The presence of pathological pleural invasion also showed a tendency toward successful detection (p = 0.081). The proposed DCNN-based algorithm yielded an accuracy of more than 90% tumor detection. These algorithms will help surgeons detect lung cancer displayed on a screen automatically.

A Small and High-Speed Driving Mechanism for 3D Shape Measurement in Monocular Endoscopy.

Three-dimensional (3D) shape acquisition has been widely introduced to enrich quantitative analysis with the combination of object shape and texture, for example, surface roughness evaluation in industry and gastrointestinal endoscopy in medicine. Shape from focus is a promising technique to measure substance surfaces in 3D space because no occlusion problem appears in principle, as does with stereo shape measurement, which is another commonly used option. We have been developing endoscopic shape measurement devices and shape reconstruction algorithms. In this paper, we propose a mechanism for driving an image sensor reciprocated for the shape from focus of 3D shape measurement in monocular endoscopy. It uses a stepping motor and a planar-end cam, which transforms the motor rotation to imaging sensor reciprocation, to implement the shape from focus of 3D shape measurement in endoscopy. We test and discuss the device in terms of its driving accuracy and application feasibility for endoscopic 3D shape measurement.

Improved segmentation of basal ganglia from MR images using convolutional neural network with crossover-typed skip connection.

PURPOSE: Accurate and automatic segmentation of basal ganglia from magnetic resonance (MR) images is important for diagnosis and treatment of various brain disorders. However, the basal ganglia segmentation is a challenging task because of the class imbalance and the unclear boundaries among basal ganglia anatomical structures. Thus, we aim to present an encoder-decoder convolutional neural network (CNN)-based method for improved segmentation of basal ganglia by focusing on skip connections that determine the segmentation performance of encoder-decoder CNNs. We also aim to reveal the effect of skip connections on the segmentation of basal ganglia with unclear boundaries. METHODS: We used the encoder-decoder CNNs with the following five patterns of skip connections: without skip connection, with full-resolution horizontal skip connection, with horizontal skip connections, with vertical skip connections, and with crossover-typed skip connections (the proposed method). We compared and evaluated the performance of the CNNs in the experiment of basal ganglia segmentation using T1-weighted MR brain images of 79 patients. RESULTS: The experimental results showed that the skip connections at each scale level help CNNs to acquire multi-scale image features, the vertical skip connections contribute on acquiring finer image features for segmentation of smaller anatomical structures with more blurred boundaries, and the crossover-typed skip connections, a combination of horizontal and vertical skip connections, provided better segmentation accuracy. CONCLUSION: This paper investigated the effect of skip connections on the basal ganglia segmentation and revealed the crossover-typed skip connections might be effective for improving the segmentation of basal ganglia with the class imbalance and the unclear boundaries.

Detecting ossification of the posterior longitudinal ligament on plain radiographs using a deep convolutional neural network: a pilot study.

BACKGROUND CONTEXT: Its rare prevalence and subtle radiological changes often lead to difficulties in diagnosing cervical ossification of the posterior longitudinal ligament (OPLL) on plain radiographs. However, OPLL progression may lead to trauma-induced spinal cord injury, resulting in severe paralysis. To address the difficulties in diagnosis, a deep learning approach using a convolutional neural network (CNN) was applied. PURPOSE: The aim of our research was to evaluate the performance of a CNN model for diagnosing cervical OPLL. STUDY DESIGN AND SETTING: Diagnostic image study. PATIENT SAMPLE: This study included 50 patients with cervical OPLL, and 50 control patients with plain radiographs. OUTCOME MEASURES: For the CNN model performance evaluation, we calculated the area under the receiver operating characteristic curve (AUC). We also compared the sensitivity, specificity, and accuracy of the diagnosis by the CNN with those of general orthopedic surgeons and spine specialists. METHODS: Computed tomography was used as the gold standard for diagnosis. Radiographs of the cervical spine in neutral, flexion, and extension positions were used for training and validation of the CNN model. We used the deep learning PyTorch framework to construct the CNN architecture. RESULTS: The accuracy of the CNN model was 90% (18/20), with a sensitivity and specificity of 80% and 100%, respectively. In contrast, the mean accuracy of orthopedic surgeons was 70%, with a sensitivity and specificity of 73% (SD: 0.12) and 67% (SD: 0.17), respectively. The mean accuracy of the spine surgeons was 75%, with a sensitivity and specificity of 80% (SD: 0.08) and 70% (SD: 0.08), respectively. The AUC of the CNN model based on the radiographs was 0.924. CONCLUSIONS: The CNN model had successful diagnostic accuracy and sufficient specificity in the diagnosis of OPLL.

Performance improvement of weakly supervised fully convolutional networks by skip connections for brain structure segmentation.

PURPOSE: For the planning and navigation of neurosurgery, we have developed a fully convolutional network (FCN)-based method for brain structure segmentation on magnetic resonance (MR) images. The capability of an FCN depends on the quality of the training data (i.e., raw data and annotation data) and network architectures. The improvement of annotation quality is a significant concern because it requires much labor for labeling organ regions. To address this problem, we focus on skip connection architectures and reveal which skip connections are effective for training FCNs using sparsely annotated brain images. METHODS: We tested 2D FCN architectures with four different types of skip connections. The first was a U-Net architecture with horizontal skip connections that transfer feature maps at the same scale from the encoder to the decoder. The second was a U-Net++ architecture with dense convolution layers and dense horizontal skip connections. The third was a full-resolution residual network (FRRN) architecture with vertical skip connections that pass feature maps between each downsampled scale path and the full-resolution scale path. The last one was a hybrid architecture with a combination of horizontal and vertical skip connections. We validated the effect of skip connections on medical image segmentation from sparse annotation based on these four FCN architectures, which were trained under the same conditions. RESULTS: For multiclass segmentation of the cerebrum, cerebellum, brainstem, and blood vessels from sparsely annotated MR images, we performed a comparative evaluation of segmentation performance among the above four FCN approaches: U-Net, U-Net++, FRRN, and hybrid architectures. The experimental results show that the horizontal skip connections in the U-Net architectures were effective for the segmentation of larger sized objects, whereas the vertical skip connections in the FRRN architecture improved the segmentation of smaller sized objects. The hybrid architecture with both horizontal and vertical skip connections achieved the best results of the four FCN architectures. We then performed an ablation study to explore which skip connections in the FRRN architecture contributed to the improved segmentation of blood vessels. In the ablation study, we compared the segmentation performance between architectures with a horizontal path (HP), an HP and vertical up paths (HP+VUPs), an HP and vertical down paths (HP+VDPs), and an HP and vertical up and down paths (FRRN). We found that the vertical up paths were effective in improving the segmentation of smaller sized objects. CONCLUSIONS: This paper investigated which skip connection architectures were effective for multiclass brain segmentation from sparse annotation. Consequently, using vertical skip connections with horizontal skip connections allowed FCNs to improve segmentation performance.

Loss Weightings for Improving Imbalanced Brain Structure Segmentation Using Fully Convolutional Networks.

Brain structure segmentation on magnetic resonance (MR) images is important for various clinical applications. It has been automatically performed by using fully convolutional networks. However, it suffers from the class imbalance problem. To address this problem, we investigated how loss weighting strategies work for brain structure segmentation tasks with different class imbalance situations on MR images. In this study, we adopted segmentation tasks of the cerebrum, cerebellum, brainstem, and blood vessels from MR cisternography and angiography images as the target segmentation tasks. We used a U-net architecture with cross-entropy and Dice loss functions as a baseline and evaluated the effect of the following loss weighting strategies: inverse frequency weighting, median inverse frequency weighting, focal weighting, distance map-based weighting, and distance penalty term-based weighting. In the experiments, the Dice loss function with focal weighting showed the best performance and had a high average Dice score of 92.8% in the binary-class segmentation tasks, while the cross-entropy loss functions with distance map-based weighting achieved the Dice score of up to 93.1% in the multi-class segmentation tasks. The results suggested that the distance map-based and the focal weightings could boost the performance of cross-entropy and Dice loss functions in class imbalanced segmentation tasks, respectively.

Complex-Valued Wavelet Spectrum Analysis of Respiratory Conditions and Its Feasibility in the Detection of Low-Functional Respiration.

Respiratory monitoring is a significant issue to reduce patient risks and medical staff labor in postoperative care and epidemic infection, particularly after the COVID-19 pandemic. Oximetry is widely used for respiration monitoring in the clinic, but it sometimes fails to capture a low-functional respiratory condition even though a patient has breathing difficulty. Another approach is breathing-sound monitoring, but this is unstable due to the indirect measurement of lung volume. Kobayashi in our team is developing a sensor measuring temporal changes in lung volume with a displacement sensor attached across the sixth and eighth ribs. For processing these respiratory signals, we propose the combination of complex-valued wavelet transform and the correlation among spectrum sequences. We present the processing results and discuss its feasibility to detect a low-functional condition in respiration. The result for detecting low-functional respiration showed good performance with a sensitivity of 0.88 and specificity of 0.88 to 1 in its receiver operating characteristic (ROC) curve.

Label cleaning and propagation for improved segmentation performance using fully convolutional networks.

PURPOSE: In recent years, fully convolutional networks (FCNs) have been applied to various medical image segmentation tasks. However, it is difficult to generate a large amount of high-quality annotation data to train FCNs for medical image segmentation. Thus, it is desired to achieve high segmentation performances even from incomplete training data. We aim to evaluate performance of FCNs to clean noises and interpolate labels from noisy and sparsely given label images. METHODS: To evaluate the label cleaning and propagation performance of FCNs, we used 2D and 3D FCNs to perform volumetric brain segmentation from magnetic resonance image volumes, based on network training on incomplete training datasets from noisy and sparse annotation. RESULTS: The experimental results using pseudo-incomplete training data showed that both 2D and 3D FCNs could provide improved segmentation results from the incomplete training data, especially by using three orthogonal annotation images for network training. CONCLUSION: This paper presented a validation for label cleaning and propagation based on FCNs. FCNs might have the potential to achieve improved segmentation performances even from sparse annotation data including possible noises by manual annotation, which can be an important clue to more efficient annotation.

Suction-fixing surgical device for assisting liver manipulation with laparoscopic forceps.

PURPOSE: Noninvasiveness and stability are significant issues in laparoscopic liver resection. Inappropriate grasping force can cause damage or serious bleeding to the liver. In addition, instability of grasping can result unsafe operations or wavered cutting. We propose a surgical device to improve stability of liver manipulation. METHODS: A proposed device adheres to the liver surface with suction fixation, then tunes its stiffness to being hard and shapes like as a bulge on the liver surface to be grasped with laparoscopic forceps. It consists of two soft beams, a chamber sponge, membrane covering the device upper, suburb extrusion wing membrane, a vacuuming tube and to-be-grasped bars. The beams are designed as being non-stretchable and easy to bend. The device is connected to a medical vacuuming pump to vacuum air in the device and then gets hard to transfer forceps operation well. This stiffness tuning mechanism by pneumatic control features the device for achieving good liver shape followability and forceps operation propagation less invasively. The proposed device was tested with rubber phantoms and porcine livers on shape followability, stiffness transition, liver invasiveness and operational usability in the experiments. RESULTS: Performance of the proposed device was assessed in experiments. The device showed good object-shape followability. It held the liver with 2.43-N force for vertical lifting and 4.90-N shear force with - 80 kPa vacuuming pressure. Invasiveness was reduced to acceptable level of liver damage. In usability test, the device grasped the liver stably and transferred surgical forceps operations to the liver surface well. CONCLUSION: The proposed device showed effective performance to improve laparoscopic liver manipulation. It held the liver stably and less invasively and transferred forceps operation force to the liver surface well.

Cardiac fiber tracking on super high-resolution CT images: a comparative study.

Purpose: High-resolution cardiac imaging and fiber analysis methods are required to understand cardiac anatomy. Although refraction-contrast x-ray CT (RCT) has high soft tissue contrast, it cannot be commonly used because it requires a synchrotron system. Microfocus x-ray CT ( µ CT ) is another commercially available imaging modality. Approach: We evaluate the usefulness of µ CT for analyzing fibers by quantitatively and objectively comparing the results with RCT. To do so, we scanned a rabbit heart by both modalities with our original protocol of prepared materials and compared their image-based analysis results, including fiber orientation estimation and fiber tracking. Results: Fiber orientations estimated by two modalities were closely resembled under the correlation coefficient of 0.63. Tracked fibers from both modalities matched well the anatomical knowledge that fiber orientations are different inside and outside of the left ventricle. However, the µ CT volume caused incorrect tracking around the boundaries caused by stitching scanning. Conclusions: Our experimental results demonstrated that µ CT scanning can be used for cardiac fiber analysis, although further investigation is required in the differences of fiber analysis results on RCT and µ CT .

Comparative analysis of surgical processes for image-guided endoscopic sinus surgery.

PURPOSE: This study proposes a method to analyze surgical performance by modeling, aligning, and comparing surgical processes. This method is intended to serve as a means to support the enhancement of surgical skills for endoscopic sinus surgeries (ESSs). We focus on surgical navigation systems used in image-guided ESSs and aim to construct a comparative analysis method for surgical processes based on the information about the surgical instruments motion obtained from the navigation system. METHODS: The proposed method consists of the following three parts: quantification of surgical features, modeling of surgical processes, and alignment and comparison of surgical process models (SPMs). First, we defined time-series parameters using the navigation-based surgical data. Second, we created SPMs by applying the defined parameters and the relative positional information of the instruments to the patient's anatomy. Third, we constructed a method to align and compare SPMs based on dynamic time warping with barycenter averaging. RESULTS: The proposed method was validated on a dataset containing surgical data obtained by an optical tracking system from 14 clinical ESS cases. We evaluated the validity of the comparative analysis by aligning and comparing SPMs between experts and residents. The validation results suggested that the proposed method could achieve proper alignment of the SPMs and clarify the differences in surgical processes between experts and residents. CONCLUSION: We developed a method to enable a time-series comparative analysis of surgical processes based on the surgical data from the navigation system. This method can allow surgeons to identify differences between their procedures and reference procedures such as experts' procedures.