HemoCountAI

• 

  Inspiration

• 

  Application Overview

• 

  App UI

• 

  Our intergration of Genimi API in the workflow

• 

  Future works

*Logo and illustrations is generated from Google Genimi

**Presentation video is narrated with AI-generated voice from Synthesia

Inspiration

The inspiration behind HemoCountAI stemmed from the need to streamline and expedite the process of blood cell analysis in healthcare settings. Traditional methods often involved manual counting, which was not only time-consuming but also prone to human error. We are also inspired to reduce the patient’s diagnosis time by adding the lab report summarization for the doctor. We aimed to harness the power of artificial intelligence to revolutionize this aspect of medical diagnostics.

What it does

HemoCountAI is a cutting-edge solution for blood cell analysis. Leveraging advanced GenAI from Google Genimi Vision Pro API. Its ability to swiftly and accurately count and classify blood cells speeds up the efficiency in medical diagnosis and treatment planning. By leveraging advanced machine learning algorithms, it provides nearly instantaneous results, which is a significant improvement over the time-consuming manual counting methods.

Moreover, the feature that processes lab reports to highlight abnormal results and provide guidance on the patient's condition is incredibly valuable. This not only saves precious time for doctors but also streamlines the diagnostic process, allowing them to focus more energy on devising the best treatment plans for their patients. With HemoCountAI, healthcare professionals can make quicker and more informed decisions, ultimately leading to better patient outcomes.

How we built it

We constructed HemoCountAI by initially utilizing a pre-trained YOLOv5 model for the detection of red blood cells (RBCs), white blood cells (WBCs), and platelets in blood cell images with mAP@0.5 score of 0.923. Subsequently, we employed Google Vision Pro to query the processed images with detections and extract the count of each blood cell type. Finally, we rendered this data as a pie chart within the application, offering users a visual representation of the analysis results.

The addition of lab report analysis is another significant advancement. Extracting text from lab reports and utilizing Genimi Pro to highlight any abnormal lab results mimics the expertise of a seasoned doctor with ten years of experience. This feature not only saves time but also enhances diagnostic accuracy, ultimately improving patient care. Overall, the combination of cutting-edge technologies and intelligent integration in HemoCountAI represents a significant leap forward in blood cell analysis and medical report analysis.

Challenges we ran into

Integrating genAI vision with Tkinter posed a significant challenge due to complexities in communication between backend algorithms and the frontend interface. Ensuring smooth image upload and synchronization between the AI backend and Tkinter frontend required careful attention and iterative testing. Selecting the appropriate visualization method for representing blood count data was another hurdle, requiring assessment of various techniques for both informativeness and user-friendliness. Through collaborative problem-solving and iterative refinement, the team successfully overcame these challenges to deliver a seamlessly integrated solution that maximized the potential of genAI vision within the Tkinter interface. The result was a user-friendly interface that effectively showcased blood cell analysis results in a visually intuitive manner, facilitating quick and accurate interpretation by healthcare professionals.

Accomplishments that we're proud of

Despite the challenges, our team successfully created HemoCountAI—a game-changer in blood cell analysis. We take pride in the speed, accuracy, and efficiency that our solution offers, providing healthcare professionals with a powerful tool to enhance patient care and diagnostics.

What we learned

Through the development of HemoCountAI, we gained invaluable insights into the potential of artificial intelligence in healthcare. We learned the importance of robust data preprocessing, model optimization, and user-centric design. Moreover, we deepened our understanding of the complexities involved in integrating AI solutions into existing healthcare infrastructure.

What's next for HemoCountAI

• This app serves a proof of concept of how Genimi can be beneficial in the space of medical laboratory.

• Looking forward, we foresee ongoing advancements and extensions for HemoCountAI. Our objective is to continuously enhance the AI model, refining it to attain superior accuracy and adaptability across various blood sample types, including a broader range of white blood cell classes such as neutrophils, eosinophils, and other types.

• Ultimately, our mission is to provide healthcare professionals with innovative tools that optimize patient care and outcomes.

• Further R&D is necessary for the app can be fully clinically approved (FDA, HIPAA compliance) to use in real-world laboratory setting.

Built With

• genimi
• matplotlib
• pillow
• pypdf2
• tkinter
• yolov5