Segmentation of tibiofemoral joint tissues from knee MRI using MtRA-Unet and incorporating shape information: Data from the Osteoarthritis Initiative

TL;DR

This work introduces MtRA-Unet, a single-stage knee MRI segmentation model that jointly learns multi-contextual features via the MRFF module and enforces anatomical structure through a Shape Reconstruction loss. Trained on the OAI/OAI-ZIB datasets, it delivers high DSC for bones ($FB$/$TB$) and cartilage ($FC$/$TC$) with a ROI-focused evaluation strategy, achieving fast inference at about $22$ seconds per subject without post-processing. The combination of MRFF and SR loss yields improved boundary and shape fidelity, particularly for smaller tissues, while ablation studies guide loss weighting, MRFF design, and data requirements. The approach shows strong potential for rapid clinical KOA assessment by providing accurate segmentation maps with minimal manual intervention and processing time.

Abstract

Knee Osteoarthritis (KOA) is the third most prevalent Musculoskeletal Disorder (MSD) after neck and back pain. To monitor such a severe MSD, a segmentation map of the femur, tibia and tibiofemoral cartilage is usually accessed using the automated segmentation algorithm from the Magnetic Resonance Imaging (MRI) of the knee. But, in recent works, such segmentation is conceivable only from the multistage framework thus creating data handling issues and needing continuous manual inference rendering it unable to make a quick and precise clinical diagnosis. In order to solve these issues, in this paper the Multi-Resolution Attentive-Unet (MtRA-Unet) is proposed to segment the femur, tibia and tibiofemoral cartilage automatically. The proposed work has included a novel Multi-Resolution Feature Fusion (MRFF) and Shape Reconstruction (SR) loss that focuses on multi-contextual information and structural anatomical details of the femur, tibia and tibiofemoral cartilage. Unlike previous approaches, the proposed work is a single-stage and end-to-end framework producing a Dice Similarity Coefficient (DSC) of 98.5% for the femur, 98.4% for the tibia, 89.1% for Femoral Cartilage (FC) and 86.1% for Tibial Cartilage (TC) for critical MRI slices that can be helpful to clinicians for KOA grading. The time to segment MRI volume (160 slices) per subject is 22 sec. which is one of the fastest among state-of-the-art. Moreover, comprehensive experimentation on the segmentation of FC and TC which is of utmost importance for morphology-based studies to check KOA progression reveals that the proposed method has produced an excellent result with binary segmentation

Figures (11)

• Figure 1: Illustration of the proposed Multi-Resolution Attentive-Unet (MtRA-Unet) architecture; Multi-Resolution Feature Fusion (MRFF) (see Figure \ref{['fig:MRFF']} module is incorporated in place of convolutions of Unet-encoder and Convolution Block Attention Module (CBAM) (see Figure \ref{['fig:CBAM']} is attached at the skip connections
• Figure 2: Schematic of the Convolutional Block Attention Module (CBAM) utilized in the paper
• Figure 3: Schematic of the proposed Multi-Resolution Feature Fusion (MRFF) Module
• Figure 4: Illustration of the proposed Shape Reconstruction (SR) loss function
• Figure 5: (a) Coronal and (b) Sagittal view of MR slices with the depiction of (I) initial and ending slices (i.e. noise-only slices with no tissue information), (II) slices with fewer occupancy of femoral and tibial bone, and (III) slices with anterior and posterior cruciate ligaments (that contain no tibia cartilage information), and (IV) medial-lateral regions (selected for segmentation analysis) for subject ID: 9471287