In order to develop a cytology support system, we first apply U-Net to separate the cytoplasm region, nucleus region, and background. After that, we extract 28 features from cell regions and a malignant judgement level was predicted by regression analysis. As a result, for isolated urothelial cells, we were able to identify malignant cells with high accuracy (area under the curve (AUC): 0.914). We also developed a system that quantitatively presents the degree of atypia, which is the basis for diagnosis, by classifying the calculated atypia into five groups according to the characteristics of malignant cells and performing statistical processing.

Panoramic X-ray images used in dentistry are taken with relatively low exposure and are a modality that can record the condition of the teeth in the patient’s mouth. In dental practice, these images are read and a patient’s medical record is prepared, but this is time-consuming for the dentist and may lead to erroneous entries. In addition, matching dental information in a database is also effective for rapid identification of bodies in the event of a large-scale disaster. Therefore, methods have been developed to automatically recognize patient information such as teeth and treatment scars from panoramic X-ray images. These methods mainly target untreated teeth for recognition, and there are still few methods that target prosthetic or root canal treated teeth. In this study, a method focusing on root canal treatment scars was added to the existing automatic tooth recognition framework, aiming to improve recognition accuracy and to create more detailed tooth charts. As a result, proposed method for the dataset including 2582 panoramic radiographs and 1740 root canals, 93.6% accuracy rate was achieved for tooth recognition by cross-validation. The precision of the root canal treatment alone was 94.6%, confirmed by the improved accuracy of the root canal treatment detection in the automatic tooth recognition.

Osteomyelitis of the jaw (OMJ) is a serious bacterial infection that affects the jaw bones, which can be caused by a variety of bacteria. Osteonecrosis of the jaw (ONJ) is bone necrosis following a bacterial infection. Symptoms may include pain, swelling, redness, fever, and bone exposure. Treatment typically involves antibiotics to kill the bacteria and surgery to remove all necrotic bone, and recent researches suggested that the early surgical resection improves the prognosis. However, it is difficult to determine the necrotic bone in pre-operative CT images. A non-invasive ONJ region recognition system is required for the precise pre-operative surgical planning, and also it has possibility of early diagnosis. This study proposes a ONJ region detection method using contrastive learning (CL) and support vector machine. The proposed method trains the CNN for feature extraction according to the CL framework, and the extracted feature is fed into support vector machine to perform two-class (normal or ONJ) classification on each subdivided image patch. In the experiment, we compared patch- and label-based CL with two types of patch extraction criteria. The proposed method was validated on 3D CT images of nine subjects, and the patch-based CL extracted ONJ region in the highest F1 of 0.734. Experimental results also suggested that the smaller stride should improve detection accuracy, and it has applicability to pixel-level ONJ detection.

This paper proposes a method to assist in iterative visual search tasks, such as those encountered in driving scenarios, by switching the focus of gaze guidance based on estimations of visual cognition. Unlike conventional methods that start assistance for the subsequent task only after the preceding one has been completed, our approach begins support for the target of the next task as soon as the visual cognition of the current task’s target is confirmed. This anticipatory gaze guidance aims to enhance task efficiency by offering assistance before the completion of the current task. During the experiments, participants were tasked with searching for specified characters in a VR environment to assess the effectiveness of the proposed method. The results indicated that for tasks with higher difficulty levels, completion times were significantly reduced with the VC-based assistance, and confirmed the effectiveness of the method.

In recent years, as online communication has become more prevalent, issues such as online harassment and cyberbullying have become increasingly common. To address these issues, it is important for commenters to objectively understand the situation of the recipient and consider the impact of their comments. In this study, we propose a system that estimates and visualizes the amount of mental damage to the comment recipient, and provides assistive messages to suppress the dissemination of offensive comments based on the estimated damage. According from experimental results, we verified that the relationship between the estimated damage and the effectiveness of the messages in suppressing the dissemination of offensive comments.

In this paper, we examine the effects of the linguistic consideration strategy used by a driver assistance robot to notify the cancellation of the automatic driving phase on reliability and annoyance. The research method was based on the linguistic consideration strategy in the politeness theory, which was designed as a reference for the phase cancellation notification utterance. A video-based experiment was conducted by crowdsourcing 240 experiment participants to subjectively evaluate the driver assistance robot’s utterances in terms of reliability and annoyance. The results showed that the “include both speaker and listener in the action,” “give reasons,” and “show respect” strategies and direct speech without linguistic consideration tended to be more reliable and less bothersome in all six stages of the automatic deactivation phase.