Embedded Files
Histopathological Image Analysis
Histopathological Image Analysis
Evaluations of surgical biopsy and postmortem tissue specimens for the diagnosis and understanding of the human disease are one of critical components of biomedical studies. Currently, these analyses are manually performed by specialists. Central problems in this process have the number and variability of disease tissue specimens, the inability to perform cross-comparison analyses, and the frequent lack of agreement between different pathologists. In addition, precise and practical integration of definitive histopathological image-based data with genetic, genomic, and clinical data is critical for precision medicine. Therefore, the need for quantitative analysis methods for classifications of disease types, subtypes, and grading the stage of the disease progression has been growing. For these goals, many recent developments in biomedical informatics use computer vision techniques for computational pathology.
Our research group is currently addressing the problem of recognizing and extracting quantitative features from Glioma histopathological images by defining, counting, and measuring morphological characteristics that can to serve as fundamental data elements capable of distinguishing key disease types and subtypes. We collaborate with Cincinnati Children’s Hospital Medical Center, OH, USA on this research.
Automatic Segmentation & Annotation for Lung Immunofluorescent Confocal Image
Automatic Segmentation & Annotation for Lung Immunofluorescent Confocal Image
Human lung development requires complex gene and cell interactions, and lung development can be studied at gene, cellular, and molecular levels. The availability of multimodal imaging data of the lung can help us visualize protein and cell localization in various lung structures. In particular, confocal Immunofluorescence (IF) images can be used in lung development modeling, and multi-class segmentation is required. However, it is difficult to get a sizable number of training images in the multi-class segmentation with recent deep learning models, e.g. the convolutional neural network (CNN) methods.
We aim to develop a system for multi-class segmentation system for lung IF images. In our laboratory, we developed a web-based annotation system for lung IF images and investigated multi-class segmentation methods. Also, we are now discussing data augmentation using deep learning approaches, for instance, GAN to solve the data imbalance problem. This research project is a part of an international research collaboration with Cincinnati Children's Hospital Medical Center, USA.
Development of Gait Assessment System Using Deep Learning for Cerebral Palsy
Development of Gait Assessment System Using Deep Learning for Cerebral Palsy
Cerebral Palsy (CP) is one of the common dysfunctions in children, and it causes motor dysfunction due to brain damage and deformities before or/and after birth. In the case of motor dysfunction, it is important to examine and evaluate a patient’s gait to measure the degree of dysfunction.
To e evaluate a patient's motor function, the Gait Deviation Index (GDI) is often used. GGI is one of the evaluation indices for gait evaluation and is well-known to reflect the grade of CP, and this index is appropriate for gait assessment of CP. Usually, GDI can be obtained by using Optical Motion Capture, but it is much burden for CP patients. For instance, we must attach a reflective marker to each joint to measure a patient’s motion. It will be much burden for CP patients. It is not suitable for accurate assessment of gait function.
On the other hand, deep learning technology is rapidly growing as computing performance has evolved, and it enables us to process a variety of data, including images. Using such technologies, we can estimate the coordinates of a patient's joint points (key points) from the given images. Such a task is called pose estimation and is being actively studied. It is assumed that the pose estimation task can be replaced with keypoints measurement with OMC. This will reduce the burden on the patient and allow anyone to obtain GDI quickly, e.g., using a camera like a smartphone.
We are now investigating to develop a method to estimate GDI using video to evaluate children with CP. Also, we are now discussing the possibilities and problems in our methods. This research project is a part of an international research collaboration with Cincinnati Children's Hospital Medical Center, USA.
(*The second image is a demonstration of a movie recording of a CP patient.)
Google Sites
Report abuse