From Algorithms to the Operating Room: Deploying Deep Learning for Automated Patient Specific Implant Design at ICML 2026

We at Meticuly are thrilled to celebrate a monumental milestone at the intersection of data science and physical medicine. On 9 July 2026, groundbreaking collaborative research spearheaded by our AI Advisor, Dr. Titipat Achakulvisut (Mahidol University), alongside Assoc. Prof. Dr. Peerapon Vateekul (Chulalongkorn University), was presented at the prestigious International Conference on Machine Learning (ICML).

The presentation, titled "Deep Learning-Based Cranial and Mandible Reconstruction for Low-Resource Clinical Deployment," showcases how advanced neural networks are completely transforming the surgical landscape. By combining cutting-edge data science with Meticuly's advanced physical manufacturing infrastructure, this research successfully bridges the gap between complex anatomical defects and flawless, custom-fit titanium solutions.

Here is an in-depth look at the innovative framework developed by Dr. Titipat and the team.

Part 1: Volumetric Cranial Reconstruction via CraNeXt

While synthetic deep learning models perform exceptionally well in controlled laboratory environments, real-world clinical data is messy. Up to 30% of real clinical scans feature severe structural obstructions around the defective area—such as bone fragments, calcifications, or remnants of old implants—which heavily distort standard reconstruction models.

Part 2: Mandible Reconstruction via the "InceptMan" Block

Jaw reconstruction following severe cancer resections or trauma is a highly complex clinical task. Rebuilding a functional jaw line utilising a fibula free flap requires intense geometric precision to restore a patient’s ability to chew, swallow, and speak.

Seamless Integration: The 3D Slicer Extension for the Patient Specific Implant

To ensure this advanced data science directly translates into clinical utility, Dr. Titipat and the team successfully integrated these machine learning modules into a fully functional 3D Slicer extension.

The resulting efficiency gains for hospitals and surgical teams are staggering:

From Hours to Minutes: Under standard manual clinical planning, completing an usable pre-operative mandible reconstruction plan takes a specialised engineer an average of 5 hours and 21 minutes. By utilising our automated zero-shot machine learning extension followed by minor adjustments, the entire planning workflow collapses to just 29 minutes and 38 seconds.

This unlocks the future potential for a single, unified deep learning system capable of generating a customised Patient-Specific Implant for virtually any anatomical structure on demand.

Meticuly is incredibly proud to back Dr. Titipat’s visionary work. By bridging academic research from Mahidol and Chulalongkorn Universities with Meticuly’s real-world industrial and regulatory capabilities, we are collectively redefining the global standard of personalised patient care.

To explore the technical specifications and data validation behind this pipeline, read the team's fully published papers across the IEEE Access and MICCAI repositories.