NerveAI: An ML-enabled Platform for Identification of Nerve Pain

Background & Unmet Need

• Headache disorders (HD) affect around 40% of the global population, around 3.1 billion people in 20211
• A large share of HD patients (>25% of HD patients in the US) also suffer from undiagnosed nerve pain, which often causes their HD
• Accurate identification of such nerve pain remains challenging due to the lack of standardized screening, delaying appropriate treatment and limiting access to care
• While some specialized providers can diagnose nerve pain through patient history and exams, most primary care physicians and general neurologists are not adequately trained to identify nerve pain
• Some HD experts use patient pain drawings to aid diagnosis, but this requires expertise in peripheral nerve anatomy, making it time-intensive, error-prone, and less accessible
• Unmet Need: Standardized screening for nerve pain among HD patients

Technology Overview

• The Technology: Platform for screening patients for nerve pain using a digital 3D model of the head on which patients draw their pain
• The platform leverages AI-based pattern recognition to automatically evaluate pain drawings to diagnose nerve pain and identify patients that are candidates for headache surgery
• A prototype of the platform has been developed and trained on 1,300 3D pain drawings
• PoC Data: The highest performing model, a multilayer perceptron (MLP) model, distinguished nerve pain from other types of head and neck pain with an AUROC of 0.879, precision of 0.943, specificity of 0.611, and sensitivity of 0.640
• Another model, XGBoost, performed exceptionally well in detecting different types of nerve pain such Trigeminal Neuralgia (AUROC: 0.954), occipital nerve pain (AUROC: 0.928), and frontal nerve pain (AUROC: 0.930)

Technology Applications

• Screen patients in specialist or non-specialist settings for nerve pain in the head and other areas of the body
• Stratify patients to non-surgical versus surgical treatment, such as nerve decompression surgery
• Predict treatment response to surgical interventions
• Differential diagnosis of nerve pain conditions, such as neuroma, thoracic outlet syndrome, sciatica, etc.

Technology Advantages

• Enables fast, inexpensive, & non-invasive screening
• Allows less specialized practitioners to assess candidacy for headache surgery
• Provides an intuitive system for patients to communicate their pain
• Early identification of nerve pain can prevent chronic pain and reduce risk of addiction to pain medication, substance abuse, and long-term disability

Figure: Rendering of mobile screening application, which allows for touch-enabled creation of pain drawings.