Back to Home

About CortexFlow

CortexFlow is a screening and exploration system engineered to demonstrate how cognitive analysis workflows can be built responsibly using modern speech-to-text models and deterministic linguistic analysis.

Strict Non-Diagnostic Disclaimer

CortexFlow is strictly a demonstration and screening exploration system. It is NOT a medical device, NOT a diagnostic tool, and MUST NOT be used for clinical diagnosis, treatment planning, or medical advice.

The cognitive metrics, longitudinal trends, and neural circuit mappings generated by this platform are experimental indicators based on linguistic and prosodic algorithms. They do not constitute a medical evaluation.

If you or a loved one are experiencing cognitive decline, memory issues, or any medical concerns, please consult a qualified healthcare professional, neurologist, or licensed medical provider immediately.

Platform Capabilities

CortexFlow analyzes speech and text inputs to extract structured, actionable data across five primary cognitive domains. Our platform is designed to provide rapid screening workflows by combining:

  • Real-Time Processing: Direct browser-based voice capture and streaming transcripts.
  • AI-Powered Transcription: Utilizing robust Gemini APIs for accurate speech-to-text conversion.
  • Deterministic Feature Extraction: Ensuring mathematical transparency over black-box AI scores.
  • Neural Mapping: Abstract metrics are visualized inside an interactive 3D brain workspace mapped to MNI152 coordinates.

Mathematical Models & Metrics

To avoid the pitfalls of opaque "AI scoring," CortexFlow relies on transparent, deterministic mathematical functions to evaluate the transcript. The system calculates deviations from standardized baselines across five neural regions.

1. Lexical Domain (Broca's Area)

Measures vocabulary richness and word retrieval capabilities.

  • Type-Token Ratio (TTR): Unique Words / Total Words. Measures lexical variation. A severe drop indicates word-finding difficulty.
  • Lexical Density: Content Words / Total Words. Content words exclude grammatical stop-words.
  • Filler Rate: The frequency of hesitation markers (e.g., "um," "uh," "like") normalized per 100 words.

2. Semantic Domain (Wernicke's Area)

Evaluates logical flow, comprehension, and tangential speech.

  • Coherence Index: Uses Jaccard Similarity across adjacent sentence vectors to measure topic adherence.
  • Idea Density: Average number of core conceptual words per utterance.
  • Tangentiality Scaling: An inverse mathematical function applied to the Coherence Index to quantify topic-wandering.

3. Prosody Domain (SMA)

Analyzes the physical timing and rhythm of speech.

  • Speech Rate: Total Words / Duration (minutes). A standard baseline is ~140 WPM.
  • Pause Frequency: The number of audio pauses (or heavy punctuation clusters) normalized per minute of speech.
  • Hesitation Ratio: The percentage of pauses exceeding 0.8 seconds relative to total pauses.

4. Syntax Domain (DLPFC)

Measures grammatical complexity and executive function in speech planning.

  • Mean Length of Utterance (MLU): Total Words / Sentence Count.
  • Clause Depth Index: A weighted sum of subordinators (e.g., "because," "although") and mid-sentence phrasal breaks.
  • Passive Voice Ratio: Pattern matching for auxiliary verbs coupled with past participles, normalized by sentence count.

5. Affective Domain (Amygdala)

Quantifies emotional resonance and certainty within the speech sample.

  • Sentiment Valence: (Positive Words - Negative Words) / (Positive + Negative + 1), scaled to a 0.0 - 1.0 index.
  • Arousal Rate: Frequency of high-intensity emotional markers.
  • Certainty Index: An inverse calculation of linguistic hedge words (e.g., "maybe," "perhaps," "guess").

Have Questions?

For inquiries regarding the mathematical models, API access, or engineering details, please reach out to the development team.

contact.manikaditya@gmail.com