AI for
visual pharmaceutical inspection
HAYAIDE is a startup developing image-based machine learning technologies to improve the efficiency and safety of dispensing inspection. By quantitatively analyzing images captured with smartphones or tablets, our system identifies pharmaceutical types and mixed states, transforming inspection workflows that have traditionally relied on visual judgment alone. We aim to reduce workload, minimize human error, and accelerate digital transformation in healthcare.
An image-based machine learning system that supports automated dispensing inspection, including tablet and powder recognition as well as mixed-state estimation.
The system operates on smartphones and tablets already available in the workplace, without requiring dedicated large-scale hardware.
HAYAIDE integrates image analysis with expertise from chemistry and healthcare to advance and standardize pharmaceutical inspection. By converting judgment processes that once depended on expert experience into reproducible digital workflows, we contribute to safer and higher-quality healthcare. Starting from Hokkaido, we aim to expand from solving local healthcare challenges to nationwide deployment.
About Us
HAYAIDE develops and implements analytical technologies that fuse image-based machine learning with knowledge from chemistry and healthcare. We are particularly focused on advancing dispensing inspection in community pharmacies and creating new systems that derive quantitative judgment from visual information. Built on academic research, our goal is to translate these technologies into practical tools that solve real challenges in healthcare settings.
MIRERUDE
Inspection of one-dose packaged medicines requires both time and sustained attention. MIRERUDE supports this process by analyzing captured images to identify tablets and powders and to estimate mixed states, helping pharmacists verify dispensing results more efficiently. This approach reduces workload while improving inspection accuracy and consistency.
- 01. Tablet and powder recognition
- 02. Support for mixed-state estimation
- 03. Operable on smartphones and tablets
- 04. Continuous model improvement
Features
Easy to deploy
Because the system runs on existing smartphones and tablets, it can be introduced into workplaces with a low implementation barrier.
Quantitative decision support
By leveraging image data for quantitative evaluation, HAYAIDE transforms inspection workflows that have traditionally depended on manual visual checks and individual experience.
Highly expandable
Through continuous model learning and improvement, the system can expand in both accuracy and scope, with future applications including inventory management and medication guidance support.
Demonstration and Validation
HAYAIDE is advancing the validation of image-based machine learning systems through demonstrations and collaborative research designed for real healthcare settings. Beyond algorithm development, we focus on implementation-ready design, including user interfaces, imaging conditions, prediction performance, and workflow integration.
Dispensing inspection support for pharmacy workflows
We are studying systems that support content and quantity checks of one-dose packaged medicines from captured images. The approach is being developed for diverse dosage forms, including tablets and powders, with the goal of reducing pharmacists' workload while improving inspection accuracy.
Quantitative image-based analytical platform
Our platform is built on research demonstrating that visual differences in powder composition and mixed states can be translated into quantitative prediction. By extending this foundation to pharmaceutical inspection, we are developing next-generation support systems that are practical for real-world use.
Get in Touch
We welcome inquiries regarding demonstration studies, collaborative research, implementation opportunities, and media coverage. Through partnerships with healthcare institutions, pharmacies, and companies, we aim to accelerate the social implementation of this technology.