About the Client
The client is an American multinational company headquartered in New York City, and one of the world’s largest pharmaceutical companies in the world, consistently ranking on the Fortune 500 list of the largest U.S. corporations.
Customer Challenge:
The client faced a critical challenge in accurately identifying diseases from X-ray reports. With an extensive database of X-ray images from patients worldwide, they realized the need to harness the power of cutting-edge technology to enhance disease detection capabilities. The sheer volume of data presented a formidable task for medical professionals to efficiently analyze and diagnose illnesses. They sought to build advanced algorithms to enhance disease detection capabilities. Their key objectives included:
- Building an Algorithm: Developing an AI-based algorithm to accurately identify diseases from a vast dataset of X-ray reports.
- Hands-on Lab: Creating an interactive, hands-on lab environment to train the model and test disease detection using X-rays from multiple patients.
- Real-time Diagnosis: Ensuring that the deep neural network algorithm could efficiently process new X-ray reports whenever a patient visited the doctor for better diagnosis.
- Automated Reports: Implementing algorithms that could generate automated reports at the end of disease discovery, streamlining the diagnosis process.
Brillio’s Solution:
Brillio assisted the client throughout the implementation in addressing their challenges and delivered an innovative and scalable solution. Brillio’s expert team then undertook meticulous data pre-processing, and, drawing on the latest advancements in deep learning, Brillio’s AI experts developed three powerful candidate classification models based on DenseNet and CNN architectures with remarkable proficiency in detecting various diseases from X-ray reports. The solution involved the following steps:
- Data science workspace setup: configured and set up the workspace, and stored an extensive dataset of 112,000 X-ray images for easy access and robust data management.
- Data Pre-processing: To optimize the model’s performance, Brillio pre-processed the data, resampling, and resizing all images to ensure uniformity and enhance the training process.
- Deep Neural Network Models: Brillio’s team of AI experts developed three candidate classification models using state-of-the-art deep neural networks. These models, based on DenseNet and CNN architectures, were selected for their ability to effectively detect various diseases from X-ray reports.
- Extensive Training: Using TensorFlow APIs and the data science workspace, Brillio meticulously trained the deep neural network models on the diverse dataset of 112,000 high-quality X-ray reports from 30,000 patients. This comprehensive training allowed the algorithms to accurately detect 15 different diseases.
Business Results:
With Brillio’s expertise, the client successfully leveraged AI to automate disease discovery from X-ray reports. The implementation of deep neural networks and the seamless integration with model output allowed for accurate disease detection, automated reporting, and scalability, revolutionizing the way diseases are diagnosed and treated. The solution stands as a testament to the potential of AI in revolutionizing healthcare and streamlining critical processes for better patient outcomes.
- Anomaly Detection: The trained algorithms achieved exceptional accuracy in identifying anomalies in X-rays, enabling healthcare professionals to detect diseases more effectively.
- Automated Report Generation: The AI-powered system automatically generated detailed reports at the end of the disease discovery process.
- Multi-Disease Detection: By developing the three powerful classification models, the solution offered the capability to detect and distinguish between multiple diseases, enhancing the diagnostic capabilities of the healthcare organization.
- Scalability: The solution was successfully scaled to process over 100,000 X-rays per day, ensuring timely and efficient diagnosis for an increasing number of patients.
- Time and Cost Optimization: Brillio’s AI solution significantly reduced the time required to detect diseases and effectively lowered the overall cost of disease identification among the 15 probable diseases, optimizing the healthcare organization’s resources.
Impact:
82-87% Disease Detection Accuracy
70% Cost Reduction in Detecting 15 Diseases
85% Reduced Time to Detect Diseases
