Analysis of psychological / physiological data during driving has been done to aim at grasping the internal state of the driver when changing from the automatic driving to the driver by awareness and attention during driving to prevent doze driving and careless driving. We confirmed that awareness and/or attention were raised by flashing the LEDs built-in in the frame of FUN'IKI glasses before passing through the signs by analyzing the eye movement during gaze. Furthermore, we conducted basic psychological experiments and confirmed whether the increase in the standard deviation of eye movements during gaze corresponded to the increase in attention.
From these results, we clarified that the movement of the FUN'IKI glasses increased the detection rate of the visual target, shortened the response time, and at the same time increased the standard deviation of the eye movement during gaze
Parameters for evaluating the swallowing function include the larynx's upward and forward amount of movement, the time from thestart of swallowing to the maximum rise, and the moving speed of the larynx. Now, only one system the Swallowing video fluoroscopicexamination of swallowing(VF) can get parameters quantitatively. But, VF has problems for patients to be exposed to radiation, and for doctorsto take so long time and so much effort. So , we can not evaluate frequently. Then, we developed the system that can evaluate quantitatively ,safety ,easier and frequently without touching and invasion. we observe the movement of the throat surface during swallowing from below, tracking ofthe Adam's apple automatically, and detect the starting of swallowing and the most rising of larynx. The Adam's apple may sink into skin withrising, so not be visible from the outside. In this cases, we tried to complement the position of the Adam's apple by lower part skin's transformation. So, we can track automatically about 90% and can get parameters for evaluating the swallowing quantitatively.
Recently, pathologists diagnose the grade of cancer by observing the pathological images. But, the number of pathologist is about 0.3 person per a unit of hospital in Japan. Moreover, the disease rate of gastric cancer is now second ranking, and the number will increase according to rapidly aging every year. A computer-aided-diagnosis is expected as the solution. In previous research, although the malignancy of gastric cancer was classified by a convolutional neural network (CNN), a sufficient classification rate was not obtained., and the effect according to dataset to CNN is not sufficiently discussed. In this paper, the malignancy of gastric cancer is classified by CNN according to Group classification. Specially, 3-class classification (Group1, Group3, Group5) is performed.