For image devices such as cameras and displays, it is difficult to obtain consistent color reproducibility due to individual manufacturer designs. We propose to improve the color reproducibility of medical images by using color charts. With the spread of COVID-19 infection, the need for telemedicine is increasing from the perspective of infection prevention. In addition, the digitization of medical images will progress in the future, and it is required that the color reproducibility of supervised data at the time of image acquisition be improved for learning / diagnosis assistance by AI. Colors in the specialized field of medical image processing are largely processed in non-standardized methods, raising many challenges in use cases where color is used for diagnostics. Smartphones are used in telemedicine, but color reproducibility may decrease due to differences in lighting as well as equipment, which may lead to a decrease in diagnostic accuracy. We have developed color charts and appropriate image processing algorithms for medical use cases. In the AI accuracy verification in pathological diagnosis, it was also confirmed that the AI diagnosis accuracy can be improved by performing image processing using a color chart.
|