Even with seemingly simple measurements, such as tumor size measurements for TNM classification, it is often difficult to get them right when you think about it properly. Clinicians often measure without thinking deeply about such things. In addition to technical (engineering) reasons, medical reasons may make appropriate measurement difficult. The principle of measurement should be studied, and black boxes are always dangerous.

The aim of our study is to develop a machine-learning (ML) based framework for estimating prostate locations with anatomical feature points (AFPs) on cone-beam computed tomography (CBCT) images for image-guided target-based patient positioning in prostate cancer radiotherapy. Three AFPs were manually determined at a “center” of prostate, a bladder contact point with prostate, and a front point of rectum on each CBCT image. ML architectures, i.e., support vector machine (SVM), artificial neural network (ANN) and random forests (RF), were incorporated into the proposed frameworks. Prostate locations were estimated using seventy-three training sets of distances between each AFP and an average prostate location with reference prostate centroids. The mean locational errors were 1.07 mm in anterior-posterior (AP) and 2.55 mm in superior-inferior (SI) direction using SVM. ANN achieved 2.48 mm in AP and 3.33 mm in SI direction, whereas RF achieved 2.16 mm in AP and 3.03 mm in SI direction, respectively. The proposed framework can be useful for imageguided target-based patient positioning.

We present a real-time hand gesture interaction system for the touchless visualization of hepatic structure in surgery, for which real-time visualization is important, particularly, during operations, but often faces the challenge of efficiently reviewing a 3D model of a patient's anatomy without requiring the surgeon to touch an input device, such as a mouse or keyboard, to maintain a sterile field. To solve this problem, we developed a touchless system based on hand gesture interaction. We used a Microsoft Kinect sensor as an input device, which performs hand detection and tracking. Our approach combines three kinds of hand states and their movements to control the visualizations of hepatic structures. Experiments demonstrate that our system can realize real-time hand gesture interactions and visualizations.

The effect of anti-VEGF therapy on macular edema due to Branch Retinal Vein Occlusion（BRVO）varies depending on patients and is therefore difficult to predict the prognosis in advance. In this paper, we present neural networks that predict LogMAR visual acuity scores improved by injecting Aflibercept from pre-treated OCT images of BRVO patients obtained before the injection. We tested two types of neural nets. The details are as follows. One neural net is a fine-tuned model whose input is only a vertical cross-sectional image of a fundus and another net is the unique CNN model with its own architecture whose input is two images : horizontal and vertical cross sections of a fundus. The training images and test images are taken using different kinds of OCT apparatuses. As a result, the fine-tuned model can predict LogMAR visual acuity scores within an error of 0.3 for 65% of the test images ? the unique CNN for 66%. These results demonstrate that both the presented nets can predict visual acuity scores even for unlearned OCT images with sufficiently high accuracy.

CT examination, which can obtain the three-dimensional information of the body, is essential in modern medicine. The three-dimensional data can be converted to two-dimensional data, and ray-sum processing is available as an application provided in image visualization software. However, the conventional ray-sum image obtained by simply projecting the CT image in the anteroposterior direction has unclear borders. In this study, we developed a novel method to generate realistic chest X-ray images from CT images. In the proposed method, CT images are projected, and a pseudo chest image is generated by nonlinear transformation, bone enhancement, unsharp masking processing, and normalization. The similarity to the actual chest X-ray images was compared by histogram analysis and visual evaluation of the pseudo chest X-ray image obtained by the proposed method and the ray-sum image obtained by the conventional method. Our results indicate that the proposed method may be effective for pseudo chest X-ray image generation using CT images.

This study aimed to evaluate the degree of variation in abdominal gradient-echo three-dimensional（3D）T1- weighted images（T1-WI）with fat suppression（FS）across various magnetic resonance imaging（MRI）systems and institutions using the same phantom. Phantom studies were performed at seven institutions using a 1.5-T MRI machine and at six institutions using a 3.0-T MRI machine. The FS rate（FSR）and coefficients of variation（CV）were calculated to evaluate uniformity of the phantom [liquid beef tallow（LBT）and saline]. The median FSR was 0.89 and 0.82 using the 1.5-T original（for routine abdominal study）and unified（unified to the extent possible）sequences and 0.91 and 0.91 using the 3.0-T original and unified sequences, respectively. The median %CV for LBT data was 10.06% and 7.99% using the 1.5-T original and unified sequences and 18.21% and 27.06% using the 3.0-T original and unified sequences, respectively. The median FSR and %CV in the central and peripheral regions in LBT data using the 1.5-T original sequence and unified sequences displayed significant differences. Conversely, the median %CV in the data using the 3.0-T original and unified sequences displayed no significant differences. The data obtained in the present multi-institutional study suggest variation in MRI data.

We determined cerebrospinal fluid (CSF) stroke volume under different intracranial pressure with individuals in the supine and sitting postures where intracranial pressure drops using an original MRI system called “Gravity MRI” that can obtain images in any posture. On a 0.4-T Gravity MRI, ECG-synchronized 2D phase contrast cine MRI technique was used to measure the flow velocity of CSF at the boundary between cranial and spinal cavities (mid-C2 level) and the cerebral aqueduct in the supine and sitting postures. We obtained the CSF stroke volume from phase-mapped velocity images over the cardiac cycle in seven healthy volunteers and compared those between supine and sitting postures. The CSF stroke volume at the mid-C2 level in the sitting posture was significantly lower than that in the supine posture,whereas the CSF stroke volume at the cerebral aqueduct did not significantly differ between both postures. There was no significant correlation in the rate of change in the CSF stroke volume at the mid-C2 level and the cerebral aqueduct. In the evaluation of intracranial pressure-regulation function using CSF pulsation, a cross-section at the mid-C2 level is more desirable than at the cerebral aqueduct.