Simplify Implant Depth Prediction as Video Grounding: A Texture Perceive Implant Depth Prediction Network

TL;DR

The paper tackles the challenge of precise dental implant depth prediction from CBCT by reframing it as a video grounding problem. It introduces TPNet, a two-component framework with an Implant Region Detector (IRD) that crops a relevant sub-volume and an Implant Depth Prediction Network (IDPNet) that regresses depth, guided by a novel Texture Perceive Loss (TPL) to capture texture variation across slices. The training objective combines $L_{reg}$, $L_{tiou}$, and $L_{TPL}$ into $L_{total}$ to jointly optimize depth accuracy and boundary alignment, while IRD reduces computation. On a dataset of 400 patients, TPNet outperforms existing methods and video grounding baselines, achieving the best Acc at high IoU and signaling potential for faster, more accurate clinical implant planning. The work advances automated dental implant guidance by leveraging texture-aware, region-cropping depth prediction.

Abstract

Surgical guide plate is an important tool for the dental implant surgery. However, the design process heavily relies on the dentist to manually simulate the implant angle and depth. When deep neural networks have been applied to assist the dentist quickly locates the implant position, most of them are not able to determine the implant depth. Inspired by the video grounding task which localizes the starting and ending time of the target video segment, in this paper, we simplify the implant depth prediction as video grounding and develop a Texture Perceive Implant Depth Prediction Network (TPNet), which enables us to directly output the implant depth without complex measurements of oral bone. TPNet consists of an implant region detector (IRD) and an implant depth prediction network (IDPNet). IRD is an object detector designed to crop the candidate implant volume from the CBCT, which greatly saves the computation resource. IDPNet takes the cropped CBCT data to predict the implant depth. A Texture Perceive Loss (TPL) is devised to enable the encoder of IDPNet to perceive the texture variation among slices. Extensive experiments on a large dental implant dataset demonstrated that the proposed TPNet achieves superior performance than the existing methods.

Figures (4)

• Figure 1: Comparison of the video grounding task and implant depth prediction task.
• Figure 2: (a) Comparison of 2D slices with different sampling intervals, $k$ represents the sampling interval. (b) Texture variation computed from the 2D slices.
• Figure 3: The architecture of the proposed texture perceive implant depth prediction framework.
• Figure 4: Detection results of TPNet trained with or without TPL loss.