RibCageImp: A Deep Learning Framework for 3D Ribcage Implant Generation

TL;DR

This work explores the feasibility of automated ribcage implant generation using deep learning using 3D U-Net architecture, and presents a framework based on 3D U-Net architecture that processes CT scans to generate patient-specific implant designs.

Abstract

The recovery of damaged or resected ribcage structures requires precise, custom-designed implants to restore the integrity and functionality of the thoracic cavity. Traditional implant design methods rely mainly on manual processes, making them time-consuming and susceptible to variability. In this work, we explore the feasibility of automated ribcage implant generation using deep learning. We present a framework based on 3D U-Net architecture that processes CT scans to generate patient-specific implant designs. To the best of our knowledge, this is the first investigation into automated thoracic implant generation using deep learning approaches. Our preliminary results, while moderate, highlight both the potential and the significant challenges in this complex domain. These findings establish a foundation for future research in automated ribcage reconstruction and identify key technical challenges that need to be addressed for practical implementation.

Figures (5)

• Figure 1: Illustration of defective ribcage $R_d$, ground truth implant $I_g$, and complete ribcage with predicted ground truth $R_d+I_g$.
• Figure 2: Data preparation for ribcage reconstruction, showing thresholding and segmentation of CT scans to obtain the defective region $R_d$ and the implant $I_g$ for model training.
• Figure 3: Training pipeline for ribcage implant generation using a 3D U-Net. The defective region $R_d$ and ground truth implant $I_g$ are processed and input into the network, which outputs the predicted implant $I_p$. The prediction is evaluated against $I_g$ with a loss function $L(I_g, I_p)$ for reconstruction accuracy.
• Figure 4: Implant prediction results on a defective ribcage, showing the ground truth and the fitted ribcage with the predicted implant using different configurations of loss terms.
• Figure 5: Comparison of ground truth and predicted implant regions showing typical failure cases (A and B) with extraneous regions and structural gaps, highlighting areas for improvement in network design and loss functions.