3D Automated Segmentation of Lower Leg Muscles Using Machine Learning on a Heterogeneous Dataset

Abstract

Quantitative MRI combines non-invasive imaging techniques to reveal alterations in muscle pathophysiology. Creating muscle-specific labels manually is time consuming and requires an experienced examiner. Semi-automatic and fully automatic methods reduce segmentation time significantly. Current machine learning solutions are commonly trained on data from healthy subjects using homogeneous databases with the same image contrast. While yielding high Dice scores (DS), those solutions are not applicable to different image contrasts and acquisitions. Therefore, the aim of our study was to evaluate the feasibility of automatic segmentation of a heterogeneous database. To create a heterogeneous dataset, we pooled lower leg muscle images from different studies with different contrasts and fields-of-view, containing healthy controls and diagnosed patients with various neuromuscular diseases. A second homogenous database with uniform contrasts was created as a subset of the first database. We trained three 3D-convolutional neuronal networks (CNN) on those databases to test performance as compared to manual segmentation. All networks, training on heterogeneous data, were able to predict seven muscles with a minimum average DS of 0.75. U-Net performed best when trained on the heterogeneous dataset (DS: 0.80 ± 0.10, AHD: 0.39 ± 0.35). ResNet and DenseNet yielded higher DS, when trained on a heterogeneous dataset (both DS: 0.86), as compared to a homogeneous dataset (ResNet DS: 0.83, DenseNet DS: 0.76). In conclusion, a CNN trained on a heterogeneous dataset achieves more accurate labels for predicting a heterogeneous database of lower leg muscles than a CNN trained on a homogenous dataset. We propose that a large heterogeneous database is needed, to make automated segmentation feasible for different kinds of image acquisitions.

Quantitative muscle MRI captures early muscle degeneration in calpainopathy

“…Muscle segmentation and tractography. Eight thigh muscles (vastus lateralis, vastus medialis, rectus femoris, semimembranosus, semitendinosus, biceps femoris, sartorius, and gracilis) and seven calf muscles (gastrocnemius medialis and lateralis, soleus, tibialis anterior, peroneus, extensor digitorum and tibialis posterior) were first segmented in patients and controls using an automated segmentation tool and subsequently optimized by an experienced rater (JF) in both legs 32 . The rater checked the automated segmentation results and manually corrected the muscle shape if necessary.…”

Robustness and stability of volume‐based tractography in a multicenter setting

“…For application in fat‐replaced muscles, highly developed segmentation frameworks are needed 12 . Recently, highly developed convolutional neural networks were shown to be feasible and accurate, but currently there is still no widespread use or availability of these methods 36–39 . These automatic approaches will revolutionize the segmentation process in the near future and facilitate mDTI assessment in clinical studies.…”

“…12 Recently, highly developed convolutional neural networks were shown to be feasible and accurate, but currently there is still no widespread use or availability of these methods. [36][37][38][39] These automatic approaches will revolutionize the segmentation process in the near future and facilitate mDTI assessment in clinical studies. Up to this point, semiautomatic approaches like the one used in this study can help to significantly reduce the segmentation time in healthy individuals.…”

Label efficient segmentation of single slice thigh CT with two-stage pseudo labels

“…With impressive performance of the deep neural network-based segmentation, Zhu et al 6 applied the H-DenseU-Net on MRI lower leg data of children with and without cerebral palsy. Rohm et al 7 created a 3D heterogeneous MRI lower leg dataset and trained a convolution network to segment muscle.…”

“…(7) Subtract the result of step (4) from step (6) to create a subcutaneous fat mask. Five coarse approximate segmentation masks are shownin Fig.2.…”