Deep Learning for Multi-Tissue Segmentation and Fully Automatic Personalized Biomechanical Models from BACPAC Clinical Lumbar Spine MRI

Artificial intelligence in musculoskeletal applications: a primer for radiologists

As an umbrella term, artificial intelligence (AI) covers machine learning and deep learning. This review aimed to elaborate on these terms to act as a primer for radiologists to learn more about the algorithms commonly used in musculoskeletal radiology. It also aimed to familiarize them with the common practices and issues in the use of AI in this domain.

Spine muscle auto segmentation techniques in MRI imaging: a systematic review

BACKGROUND

The accurate segmentation of spine muscles plays a crucial role in analyzing musculoskeletal disorders and designing effective rehabilitation strategies. Various imaging techniques such as MRI have been utilized to acquire muscle images, but the segmentation process remains complex and challenging due to the inherent complexity and variability of muscle structures. In this systematic review, we investigate and evaluate methods for automatic segmentation of spinal muscles.

METHODS

Data for this study were obtained from PubMed/MEDLINE databases, employing a search methodology that includes the terms 'Segmentation spine muscle' within the title, abstract, and keywords to ensure a comprehensive and systematic compilation of relevant studies. Systematic reviews were not included in the study.

RESULTS

Out of 369 related studies, we focused on 12 specific studies. All studies focused on segmentation of spine muscle use MRI, in this systematic review subjects such as healthy volunteers, back pain patients, ASD patient were included. MRI imaging was performed on devices from several manufacturers, including Siemens, GE. The study included automatic segmentation using AI, segmentation using PDFF, and segmentation using ROI.

CONCLUSION

Despite advancements in spine muscle segmentation techniques, challenges still exist. The accuracy and precision of segmentation algorithms need to be improved to accurately delineate the different muscle structures in the spine. Robustness to variations in image quality, artifacts, and patient-specific characteristics is crucial for reliable segmentation results. Additionally, the availability of annotated datasets for training and validation purposes is essential for the development and evaluation of new segmentation algorithms. Future research should focus on addressing these challenges and developing more robust and accurate spine muscle segmentation techniques to enhance clinical assessment and treatment planning for musculoskeletal disorders.

Neural network segmentation of disc volume from magnetic resonance images and the effect of degeneration and spinal level

Background

Magnetic resonance imaging (MRI) noninvasively quantifies disc structure but requires segmentation that is both time intensive and susceptible to human error. Recent advances in neural networks can improve on manual segmentation. The aim of this study was to establish a method for automatic slice-wise segmentation of 3D disc volumes from subjects with a wide range of age and degrees of disc degeneration. A U-Net convolutional neural network was trained to segment 3D T1-weighted spine MRI.

Methods

Lumbar spine MRIs were acquired from 43 subjects (23-83 years old) and manually segmented. A U-Net architecture was trained using the TensorFlow framework. Two rounds of model tuning were performed. The performance of the model was measured using a validation set that did not cross over from the training set. The model version with the best Dice similarity coefficient (DSC) was selected in each tuning round. After model development was complete and a final U-Net model was selected, performance of this model was compared between disc levels and degeneration grades.

Results

Performance of the final model was equivalent to manual segmentation, with a mean DSC = 0.935 ± 0.014 for degeneration grades I-IV. Neither the manual segmentation nor the U-Net model performed as well for grade V disc segmentation. Compared with the baseline model at the beginning of round 1, the best model had fewer filters/parameters (75%), was trained using only slices with at least one disc-labeled pixel, applied contrast stretching to its input images, and used a greater dropout rate.

Conclusion

This study successfully trained a U-Net model for automatic slice-wise segmentation of 3D disc volumes from populations with a wide range of ages and disc degeneration. The final trained model is available to support scientific use.

Deep learning-based structure segmentation and intramuscular fat annotation on lumbar magnetic resonance imaging

Background

Lumbar disc herniation (LDH) is a prevalent cause of low back pain. LDH patients commonly experience paraspinal muscle atrophy and fatty infiltration (FI), which further exacerbates the symptoms of low back pain. Magnetic resonance imaging (MRI) is crucial for assessing paraspinal muscle condition. Our study aims to develop a dual-model for automated muscle segmentation and FI annotation on MRI, assisting clinicians evaluate LDH conditions comprehensively.

Methods

The study retrospectively collected data diagnosed with LDH from December 2020 to May 2022. The dataset was split into a 7:3 ratio for training and testing, with an external test set prepared to validate model generalizability. The model's performance was evaluated using average precision (AP), recall and F1 score. The consistency was assessed using the Dice similarity coefficient (DSC) and Cohen's Kappa. The mean absolute percentage error (MAPE) was calculated to assess the error of the model measurements of relative cross-sectional area (rCSA) and FI. Calculate the MAPE of FI measured by threshold algorithms to compare with the model.

Results

A total of 417 patients being evaluated, comprising 216 males and 201 females, with a mean age of 49 ± 15 years. In the internal test set, the muscle segmentation model achieved an overall DSC of 0.92 ± 0.10, recall of 92.60%, and AP of 0.98. The fat annotation model attained a recall of 91.30%, F1 Score of 0.82, and Cohen's Kappa of 0.76. However, there was a decrease on the external test set. For rCSA measurements, except for longissimus (10.89%), the MAPE of other muscles was less than 10%. When comparing the errors of FI for each paraspinal muscle, the MAPE of the model was lower than that of the threshold algorithm.

Conclusion

The models demonstrate outstanding performance, with lower error in FI measurement compared to thresholding algorithms.

Deep learning assisted segmentation of the lumbar intervertebral disc: a systematic review and meta-analysis

BACKGROUND

In recent years, deep learning (DL) technology has been increasingly used for the diagnosis and treatment of lumbar intervertebral disc (IVD) degeneration. This study aims to evaluate the performance of DL technology for IVD segmentation in magnetic resonance (MR) images and explore improvement strategies.

METHODS

We developed a PRISMA systematic review protocol and systematically reviewed studies that used DL algorithm frameworks to perform IVD segmentation based on MR images published up to April 10, 2024. The Quality Assessment of Diagnostic Accuracy Studies-2 tool was used to assess methodological quality, and the pooled dice similarity coefficient (DSC) score and Intersection over Union (IoU) were calculated to evaluate segmentation performance.

RESULTS

45 studies were included in this systematic review, of which 16 provided complete segmentation performance data and were included in the quantitative meta-analysis. The results indicated that DL models showed satisfactory IVD segmentation performance, with a pooled DSC of 0.900 (95% confidence interval [CI]: 0.887-0.914) and IoU of 0.863 (95% CI: 0.730-0.995). However, the subgroup analysis did not show significant effects of factors on IVD segmentation performance, including network dimensionality, algorithm type, publication year, number of patients, scanning direction, data augmentation, and cross-validation.

CONCLUSIONS

This study highlights the potential of DL technology in IVD segmentation and its further applications. However, due to the heterogeneity in algorithm frameworks and result reporting of the included studies, the conclusions should be interpreted with caution. Future research should focus on training generalized models on large-scale datasets to enhance their clinical application.

A novel CNN-based image segmentation pipeline for individualized feline spinal cord stimulation modeling

Objective. Spinal cord stimulation (SCS) is a well-established treatment for managing certain chronic pain conditions. More recently, it has also garnered attention as a means of modulating neural activity to restore lost autonomic or sensory-motor function. Personalized modeling and treatment planning are critical aspects of safe and effective SCS (Rowald and Amft 2022 Front. Neurorobotics 16 983072, Wagneret al2018 Nature 563 65-71). However, the generation of spine models at the required level of detail and accuracy requires time and labor intensive manual image segmentation by human experts. This study aims to develop a maximally automated segmentation routine capable of producing high-quality anatomical models, even with limited data, to facilitate safe and effective personalized SCS treatment planning.Approach. We developed an automated image segmentation and model generation pipeline based on a novel convolutional neural network (CNN) architecture trained on feline spinal cord magnetic resonance imaging data. The pipeline includes steps for image preprocessing, data augmentation, transfer learning, and cleanup. To assess the relative importance of each step in the pipeline and our choice of CNN architecture, we systematically dropped steps or substituted architectures, quantifying the downstream effects in terms of tissue segmentation quality (Jaccard index and Hausdorff distance) and predicted nerve recruitment (estimated axonal depolarization).Main results. The leave-one-out analysis demonstrated that each pipeline step contributed a small but measurable increment to mean segmentation quality. Surprisingly, minor differences in segmentation accuracy translated to significant deviations (ranging between 4% and 13% for each pipeline step) in predicted nerve recruitment, highlighting the importance of careful workflow design. Additionally, transfer learning techniques enhanced segmentation metric consistency and allowed generalization to a completely different spine region with minimal additional training data.Significance. To our knowledge, this work is the first to assess the downstream impacts of segmentation quality differences on neurostimulation predictions. It highlights the role of each step in the pipeline and paves the way towards fully automated, personalized SCS treatment planning in clinical settings.

Thresholding approaches for estimating paraspinal muscle fat infiltration using T1- and T2-weighted MRI: Comparative analysis using water-fat MRI

Background

Paraspinal muscle fat infiltration is associated with spinal degeneration and low back pain, however, quantifying muscle fat using clinical magnetic resonance imaging (MRI) techniques continues to be a challenge. Advanced MRI techniques, including chemical-shift encoding (CSE) based water-fat MRI, enable accurate measurement of muscle fat, but such techniques are not widely available in routine clinical practice.

Methods

To facilitate assessment of paraspinal muscle fat using clinical imaging, we compared four thresholding approaches for estimating muscle fat fraction (FF) using T1- and T2-weighted images, with measurements from water-fat MRI as the ground truth: Gaussian thresholding, Otsu's method, K-mean clustering, and quadratic discriminant analysis. Pearson's correlation coefficients (r), mean absolute errors, and mean bias errors were calculated for FF estimates from T1- and T2-weighted MRI with water-fat MRI for the lumbar multifidus (MF), erector spinae (ES), quadratus lumborum (QL), and psoas (PS), and for all muscles combined.

Results

We found that for all muscles combined, FF measurements from T1- and T2-weighted images were strongly positively correlated with measurements from the water-fat images for all thresholding techniques (r = 0.70-0.86, p < 0.0001) and that variations in inter-muscle correlation strength were much greater than variations in inter-method correlation strength.

Conclusion

We conclude that muscle FF can be quantified using thresholded T1- and T2-weighted MRI images with relatively low bias and absolute error in relation to water-fat MRI, particularly in the MF and ES, and the choice of thresholding technique should depend on the muscle and clinical MRI sequence of interest.

Accurate Intervertebral Disc Segmentation Approach Based on Deep Learning

Automatically segmenting specific tissues or structures from medical images is a straightforward task for deep learning models. However, identifying a few specific objects from a group of similar targets can be a challenging task. This study focuses on the segmentation of certain specific intervertebral discs from lateral spine images acquired from an MRI scanner. In this research, an approach is proposed that utilizes MultiResUNet models and employs saliency maps for target intervertebral disc segmentation. First, a sub-image cropping method is used to separate the target discs. This method uses MultiResUNet to predict the saliency maps of target discs and crop sub-images for easier segmentation. Then, MultiResUNet is used to segment the target discs in these sub-images. The distance maps of the segmented discs are then calculated and combined with their original image for data augmentation to predict the remaining target discs. The training set and test set use 2674 and 308 MRI images, respectively. Experimental results demonstrate that the proposed method significantly enhances segmentation accuracy to about 98%. The performance of this approach highlights its effectiveness in segmenting specific intervertebral discs from closely similar discs.

K2S Challenge: From Undersampled K-Space to Automatic Segmentation

Magnetic Resonance Imaging (MRI) offers strong soft tissue contrast but suffers from long acquisition times and requires tedious annotation from radiologists. Traditionally, these challenges have been addressed separately with reconstruction and image analysis algorithms. To see if performance could be improved by treating both as end-to-end, we hosted the K2S challenge, in which challenge participants segmented knee bones and cartilage from 8× undersampled k-space. We curated the 300-patient K2S dataset of multicoil raw k-space and radiologist quality-checked segmentations. 87 teams registered for the challenge and there were 12 submissions, varying in methodologies from serial reconstruction and segmentation to end-to-end networks to another that eschewed a reconstruction algorithm altogether. Four teams produced strong submissions, with the winner having a weighted Dice Similarity Coefficient of 0.910 ± 0.021 across knee bones and cartilage. Interestingly, there was no correlation between reconstruction and segmentation metrics. Further analysis showed the top four submissions were suitable for downstream biomarker analysis, largely preserving cartilage thicknesses and key bone shape features with respect to ground truth. K2S thus showed the value in considering reconstruction and image analysis as end-to-end tasks, as this leaves room for optimization while more realistically reflecting the long-term use case of tools being developed by the MR community.