Low-Cost, Open-Source Design and FDM Fabrication of Patient-Specific Orthopedic Forearm Casts from Consumer RGB-D Scanning

Conventional plaster and fiberglass forearm casts are formed by hand from wet material that hardens in situ , making internal geometry neither objectively measurable nor reproducible. This article presents a fully open-source design-to-fabrication pipeline that produces a patient-specific, Voronoi-ventilated forearm cast for FDM additive manufacturing directly from a 36-frame consumer RGB-D depth scan. An Intel RealSense D435 depth camera paired with a Raspberry Pi 4 captures aligned depth-color frame pairs as the forearm rotates on a manual turntable at 10° increments. Truncated Signed Distance Function (TSDF) volumetric fusion at 3 mm voxel resolution reconstructs the arm surface; a two-stage voxel-regularization followed by Screened Poisson reconstruction guarantees a watertight mesh topology (Euler characteristic χ = 2 ) required for Boolean solid modeling. A cast shell with a 3 mm anatomical clearance gap and a 4 mm structural wall is generated automatically. One hundred and fifty Voronoi-distributed, surface-normal-aligned ventilation holes cover approximately 35% of the shell area. A clamshell split with fastener alignment enables tool-free donning and doffing. The entire pipeline executes within a Google Colab notebook. Controlled synthetic geometric validation across five swept-ellipse forearm-like geometries spanning pediatric-to-adult size variants yielded mean inner-surface clearance of 2.998–2.999 mm against the 3 mm design target, with preliminary Finite Element Method (FEM) screening on a synthetic cast shell indicating low displacement (0.0041 mm) and stress (0.100 MPa p95 von Mises) under a representative 50 N load. A physical PLA prototype confirmed the dimensional accuracy of the generated STereoLithography (STL) geometry. The complete scanning station costs approximately $200 USD; per-cast FDM material cost is $22, a 95.1% reduction relative to the nearest commercial alternative ($450). End-to-end processing time was under 10 min on the free Google Colab tier. The pipeline is deterministic across repeated runs, and the complete code and validation artifacts are publicly available at https://github.com/GZEN24/SCAN23D .