Real-Time Anomaly Detection in Medical Images using Embedded Deep Learning Models on iOS and Android Devices

Data Sources

Images used to train deep learning models for real time anomaly detection from medical images were obtained from the following sources for non-commercial illustrations.

Building the Deep Learning Models for Real Time Detection of Anomalies

Deep learning models were created using Google TensorFlow. They were subjected to further processing and compression to produce TFLite version of the models. TFLite models were implanted onto the Flutter mobile app and deployed to iOS and Android devices.

Real Time Medical AI at your Fingertips Demos

Demo Part 1 - Brain MRI

Detect specific anomalies based on location of the lesions:

• Glioma - deep in the brain
• Meningioma - closer to the skull
• Normal

Demo Part 2 - Thoracic Histopathology

Detect difference in cellular morphology:

• Adenocarcinoma
• Squamous Cell Carcinoma
• Normal

Demo Part 3 - Histopathology of Osteosarcoma

Detect the presence of lesions:

• Benign
• Malignant
• Normal

Benefits

• Regardless of your objectives and perspective, just leave it to the machine learning models
• Your mobile phone is your most convenient tool
• So many cases, so little time, so few experts, use AI on your mobile for triage - quick and dirty way
• The power of AI is now at your fingertips