CatScan - August Diary

August marks the pivotal month when the Cat-Scan app takes its final shape. The focus shifts from coding individual features to integrating the app, preparing for user testing, and fine-tuning the database and user interface. By the end of this month, the core functionalities—image capture, pupil detection, and AI predictions—are up and running, ready for doctor testing in September.

The month begins with the team working on integrating all the pieces that were developed earlier. The app now has a full image capture and processing flow. Students confirm that the cropping tool works, and a new feature is added to allow the cropping ring to shift, making it adaptable for off-center pupils. This was one of the suggestions that came up in earlier meetings, and it addresses one of the main challenges—ensuring accurate pupil detection in real-world, non-ideal conditions. The team also focuses on the functionality of saving image names and their associated IDs in the database, which is crucial for tracking and categorizing cataract images for machine learning purposes.

In parallel, the database receives attention. The images, along with their metadata (such as patient ID and cataract status), are stored properly. The team ensures that each image is linked to a unique identifier, making it easier to manage the data as the app scales. A critical task is setting up the database to handle a large volume of images—an essential consideration for this tool’s use in rural areas where many patients will need to be screened.

The app is also tested for user experience. The interface has been designed with simplicity in mind, offering easy-to-follow instructions for volunteers. The next step is to conduct internal testing, which involves capturing pupil images, saving them in the database, and running the AI model to predict cataract status. The goal is to complete this internal testing phase before the doctor testing scheduled for the third week of September.

Next Tasks (for September):

• Complete internal testing of the full app workflow (image capture, prediction, and database save).
• Address any UI issues and refine the database integration to improve speed and reliability.
• Test the app with a small group of doctors to gather feedback and identify areas for improvement.
• Finalize machine learning model training using the captured pupil images.

Alignment with May and June Diaries

This August diary aligns well with the goals set out in May and June. May focused on understanding the problem, creating a detailed workflow, and setting up the necessary tools for data capture and processing. June continued the work with data cleaning and model pipeline creation, leading up to the design of a working capture system and the initial app prototype. August, as outlined here, completes the integration of these components, moving the project closer to real-world testing with doctors and improving the system for actual deployment.