Time-Lapse Macro Imaging and LLM-Assisted Phenotypic Analysis of Release Dynamics in Nifedipine Controlled-Release Tablets.

PURPOSE: To develop an exploratory flow-through time-lapse macro-imaging framework combining rule-based image analysis and multimodal large language models (LLMs) for comparing release phenotypes of nifedipine osmotic pump tablets.

METHODS

Sixteen marketed products, comprising Adalat OROS and 15 generics, were imaged in triplicate for 36 h under sequential pH 1.0 (2 h) and pH 6.8 (34 h) conditions at 37 ± 0.5°C. Images acquired every 30 s were analyzed using a pixel color-based drug similarity method (PCDS), the Dis-GPT agent, and a pharmaceutical knowledge-base-assisted DIS Chatflow.

RESULTS

The framework captured drug-layer extrusion, local aggregation, release continuity, and push-layer morphology. Eight generics (02, 03, 04, 10, 12, 13, 15, and 16) were classified as similar to the reference by all three methods. Product 08 was consistently classified as different, exhibiting granular extrusion, bead-like aggregation, and reduced continuity; products 06 and 07 showed method-dependent deviations in release timing and push-layer extrusion.

CONCLUSIONS

Flow-through time-lapse imaging revealed process-level release features not represented by cumulative dissolution profiles. Combining transparent pixel-based measurements with structured LLM analysis provides an exploratory framework for comparing visible release phenotypes and generating hypotheses about release mechanisms in controlled-release tablets.