Shimadzu had 13 patents in artificial intelligence during Q3 2023. These include a training support method for chromatograms, a sensory test and entrusted analysis method, a method for analyzing food samples and predicting examination results, a method for automatically detecting degradation in document search performance, and a method for calculating the certainty factor of peak picking using semantic segmentation technology. GlobalData’s report on Shimadzu gives a 360-degreee view of the company including its patenting strategy. Buy the report here.
Shimadzu grant share with artificial intelligence as a theme is 46% in Q3 2023. Grant share is based on the ratio of number of grants to total number of patents.
Application: Training support method (Patent ID: US20230267312A1)
The patent filed by Shimadzu Corp. describes a training support method for assisting in the training of an estimation model used to detect peaks in a signal waveform acquired by an analysis device. The method involves acquiring a first signal waveform from the analysis device and displaying it on a display device. A second signal waveform, similar to the first waveform, and second peak information specifying the peaks of the second waveform are acquired from a storage device that stores annotated signals. The second waveform and the second peak information are displayed on the display device. The user then inputs first peak information specifying the peaks of the first waveform, and the estimation model is trained based on the first waveform and the first peak information.
The method includes calculating a similarity degree between the first waveform and each of the storage signal waveforms in the storage device. The storage signal waveform with a similarity degree greater than or equal to a threshold is acquired as the second waveform. The user inputs the first peak information after the first waveform and the second waveform with the second peak information are displayed on the display device.
Alternatively, the second waveform and the second peak information can be acquired after the user inputs the first peak information. In this case, the method involves calculating both the similarity degree between the first waveform and the storage signal waveforms and the similarity degree between the first peak information and the storage peak information. The second waveform is acquired if the first similarity degree is greater than or equal to a first threshold and the second similarity degree is greater than or equal to a second threshold.
The method also includes displaying a first peak information image indicating the first peak information on the display device. Additionally, a detected peak information image indicating the peaks of the first waveform detected using the estimation model can be displayed. The user can input the first peak information after the detected peak information image and the first waveform are displayed.
Furthermore, the user can input a first point and a second point while the first waveform is displayed. The second peak information image is a line image, and the region surrounded by the second waveform and the line image indicates the peaks of the second waveform. The peak region specified by the first peak information is a region surrounded by a line connecting the first point and the second point and a line included in the first waveform.