Enhancing Knee X-ray Diagnostics with AI
Kavya_Mohan |
0
Description:
Osteoarthritis affects approximately 520 million people globally, with many cases undiagnosed due to delays and inaccuracies in traditional knee X-ray interpretation. These diagnostic challenges contribute to poor patient outcomes and increased morbidity. This talk focuses on how we can leverage Python and AI to enhance the quality, effectiveness, and efficiency of knee X-ray diagnostics. The X-ray reporting field faces a severe shortage of radiologists, especially in high-demand regions like India. Consequently, clinicians often need to interpret X-rays without specialist input, potentially affecting the accuracy and effectiveness of patient care. Our study evaluates the impact of a Python-based AI-enhanced diagnostic process on knee X-ray interpretations. Using quality improvement methodology, we applied iterative enhancements to a Faster-RCNN-based AI model. The model, integrated into the diagnostic process of a large imaging network, was trained on 73,456 labelled knee X-ray images to detect 12 common pathologies that account for 88.2% of abnormal studies.
Outline of the Talk:
Problem Overview: Diagnostic challenges in knee X-rays and the impact on patient care.
DICOM: Overview of Pydicom and Basic Medical Image Processing Using Python.
Building the AI Model: Overview of Faster-RCNN and its implementation in Python.
-- Data Preparation: Techniques for Handling and Labelling Large Datasets Using FastAPI.
-- Model Training: Using PyTorch to train the AI model.
-- Evaluation Metrics: Accuracy, precision, recall, and F1 score.
Integrating AI into Diagnostic Processes: Deploying the model using Flask and Docker.
Addressing Challenges: Ensuring data integrity and managing technical issues.
Demo: Live demonstration of the AI model in action.
Future Work: Expanding the study to multiple regions and additional pathologies.
Prerequisites:
Basic understanding of machine learning and Python programming.
Speaker Info:
Kavya Mohan is a Data Scientist at 5C Network, specialising in AI-driven solutions for medical diagnostics using Python. Passionate about the intersection of medical science and technology, she has played a key role in developing AI models for CT scans, X-rays, and MRIs. Although her background is in Electronics and Instrumentation Engineering, Kavya self-taught Python and transitioned into data science, where she leverages libraries such as FastAI, Pytorch and Pydicom to create advanced medical imaging solutions. Her work highlights the transformative power of Python in healthcare.