![SonomaEye-full-color-vector[2].png](https://static.wixstatic.com/media/bed00b_da20b9cc704044b0b048ac8ef9190470~mv2_d_3024_2195_s_2.png/v1/fill/w_177,h_123,al_c,q_85,usm_0.66_1.00_0.01,enc_avif,quality_auto/SonomaEye-full-color-vector%5B2%5D.png)
Thank you for your submission!
The Gene Therapy Blind Spot: Video-Based Volumetric Analysis (V3xV) for Precision Subretinal Dosing
David Almeida MD MBA PhD
Presenter:
David Almeida MD MBA PhD, Michael Singer MD, Vinit Mahajan MD PhD
Authors:
Affiliation:
1. The Centers for Advanced Surgical Exploration (CASEx)
PURPOSE: Subretinal gene therapy delivery requires precise micro-dosing to ensure therapeutic efficacy and minimize toxicity. The current standard of care relies on the surgeon’s subjective visual estimation of bleb size, which is prone to significant variability. This dosing blind spot compromises clinical trial data integrity and patient safety. We present the validation of V3xV (SurgVECTORai), a novel artificial intelligence (AI) platform capable of real-time, automated volumetric analysis of subretinal blebs using standard surgical video feeds.
METHODS: We utilized the SurgTECH.ai proprietary dataset, comprising high-resolution vitreoretinal surgical videos synchronized with intraoperative OCT (iOCT). The V3xV computer vision pipeline was trained using a frame-by-frame segmentation approach. The algorithm executes three parallel functions: 1) Locate: AI classification of the injection site (subretinal vs. intravitreal) to detect reflux; 2) Measure: Geometric characterization of the bleb using an ellipsoid approximation model derived from pixel-based dimension estimation; and 3) Guide: Generation of "Go/No-Go" safety alerts based on flow rate and volume thresholds. The system was validated against manual volumetric measurements confirmed by iOCT.
RESULTS: Analysis demonstrated that subjective visual estimation by surgeons resulted in a mean dosing error of 32 microliters compared to the target dose. In contrast, the V3xV automated analysis achieved a dosing precision of 0.1 microliters. Real-time processing enabled the generation of an objective Dosing Log, documenting total volume infused, flow rate consistency, and bleb retention, which was unavailable via standard surgical recording.
CONCLUSIONS: V3xV represents a paradigm shift from subjective estimation to objective quantification in gene therapy delivery. By eliminating the dosing blind spot, this technology offers a scalable method to standardize surgical delivery across multi-center clinical trials. Integration of V3xV into surgical workflows can reduce protocol deviations, enhance the safety profile of subretinal injections, and potentially improve the correlation between administered dose and therapeutic effect.
Attachments:
