LST-AI: a Deep Learning Ensemble for Accurate MS Lesion Segmentation.
Tun WiltgenJulian McGinnisSarah SchlaegerCuiCi VoonAchim BertheleDaria BischlLioba GrundlNikolaus WillMarie MetzDavid SchinzDominik SeppPhilipp PruckerBenita Schmitz-KoepClaus ZimmerBjoern MenzeDaniel RueckertBernhard HemmerJan KirschkeMark MühlauBenedikt WiestlerPublished in: medRxiv : the preprint server for health sciences (2023)
Automated segmentation of brain white matter lesions is crucial for both clinical assessment and scientific research in multiple sclerosis (MS). Over a decade ago, we introduced a lesion segmentation tool, LST, engineered with a lesion growth algorithm (LST-LGA). While recent lesion segmentation approaches have leveraged artificial intelligence (AI), they often remain proprietary and difficult to adopt. Here, we present LST-AI, an advanced deep learning-based extension of LST that consists of an ensemble of three 3D-UNets. LST-AI specifically addresses the imbalance between white matter (WM) lesions and non-lesioned WM. It employs a composite loss function incorporating binary cross-entropy and Tversky loss to improve segmentation of the highly heterogeneous MS lesions. We train the network ensemble on 491 MS pairs of T1w and FLAIR images, collected in-house from a 3T MRI scanner, and expert neuroradiologists manually segmented the utilized lesion maps for training. LST-AI additionally includes a lesion location annotation tool, labeling lesion location according to the 2017 McDonald criteria (periventricular, infratentorial, juxtacortical, subcortical). We conduct evaluations on 270 test cases -comprising both in-house (n=167) and publicly available data (n=103)-using the Anima segmentation validation tools and compare LST-AI with several publicly available lesion segmentation models. Our empirical analysis shows that LST-AI achieves superior performance compared to existing methods. Its Dice and F1 scores exceeded 0.5, outperforming LST-LGA, LST-LPA, SAMSEG, and the popular nnUNet framework, which all scored below 0.45. Notably, LST-AI demonstrated exceptional performance on the MSSEG-1 challenge dataset, an international WM lesion segmentation challenge, with a Dice score of 0.65 and an F1 score of 0.63-surpassing all other competing models at the time of the challenge. With increasing lesion volume, the lesion detection rate rapidly increased with a detection rate of >75% for lesions larger than 60mm 3 . Given its higher segmentation performance, we recommend that research groups currently using LST-LGA transition to LST-AI. To facilitate broad adoption, we are releasing LST-AI as an open-source model, available as a command-line tool, dockerized container, or Python script, enabling diverse applications across multiple platforms.
Keyphrases
- deep learning
- artificial intelligence
- convolutional neural network
- big data
- multiple sclerosis
- white matter
- machine learning
- mass spectrometry
- magnetic resonance imaging
- ms ms
- high resolution
- computed tomography
- clinical practice
- high throughput
- single cell
- rna seq
- subarachnoid hemorrhage
- network analysis
- ionic liquid
- optical coherence tomography
- blood brain barrier
- contrast enhanced