Disease Screening in Healthcare

Industry Challenges & Trends

Healthcare providers face ever-increasing costs of disease diagnosis due to the growing demand for screening tests and the shortage of trained doctors amidst preventive healthcare awareness growth. In fact, more than 97% of those screened have no potential problems identified but require almost the same level of care by a trained doctor. The inability to replicate the routine screening process at scale shifts the burden to healthcare providers, affecting their operational efficiency and limiting patients’ access to affordable healthcare.

The implementation of AI for computer-aided medical image analysis and pathology detection enables doctors to diagnose faster and more accurately, thereby allowing to serve more patients, increase recovery rate, reduce screening costs, and generate financial opportunity for healthcare providers.

Disease Screening Solution for Automated Patient Diagnosis

The Provectus Disease Screening Solution applies an AI-driven image analysis and anomaly detection engine to process, analyze, and label a wide range of medical images, including eye and skin screens, MRIs, X-ray, tomography and mammography images.

The solution is adjusted to accurately detect pathology in images, also to identify normal “healthy” screens. This enables physicians not only to diagnose faster, but also to spend more time on scrutinizing abnormal screens in the disputable cases, thereby continuously improving the accuracy of the system. By sifting out healthy patients, the system allows healthcare providers to eliminate up to 90% of related triage costs.

The solution is designed to be implemented selectively to mitigate technology and data privacy concerns of a full-scale AI/ML adoption across the healthcare organization.

Real-world Case

Pr3vent is a medical technology company that provides an AI-driven eye screening system to diagnose eye pathology in healthy term infants. Powered by Disease Screening Solution, the system enables physicians to diagnose eye conditions faster and more accurately, allowing for early treatment to prevent the development of vision loss. Making eye screening accessible to four million babies in the US alone, it can potentially save the society up to $3 billion per year.

Key Features

• Pre-built machine learning architecture, including infrastructure for model design, training, and deployment in production
• Internal labeling system to create and customize datasets, with a user-friendly, web-based UI for physicians
• Pre-trained machine learning algorithms that learn, adapt, and improve disease detection and classification accuracy based on data
• Cloud vendor agnostic — The architecture is certified to be used with different cloud providers

How It Works

The Provectus Disease Screening Solution consists of three components:

1. Image labeling tool
2. ML model building and training tool
3. UI for disease screening application

The image labeling tool helps medical professionals process and label examination images. With a user-friendly UI accessible via the Web, the tool allows doctors to upload a wide variety of medical images to label them as either normal or abnormal or to drop low-quality images from the tool.

The ML model building and training tool apply the dataset generated in the image labeling tool to create and train ML models. The dataset can be extended and enhanced by adding more labeled images. ML models with the highest prediction accuracy are selected for use in the main application, which is used by physicians on site.

Physicians access the main app’s UI to check how accurately pathologies have been detected and classified by applied ML models. To simplify the assessment of diagnostics results, the application displays the following details to the physician: Label — either normal or abnormal; Title and Version of the applied ML model(s); Prediction; Prediction accuracy; and Explainability section showcasing pathology areas in a given image.