Deep learning based tractography with TractSeg in patients with hemispherotomy: Evaluation and refinement

IF 3.4 2区医学 Q2 NEUROIMAGING

Neuroimage-Clinical Pub Date : 2025-01-01 DOI:10.1016/j.nicl.2025.103738

Johannes Gruen , Tobias Bauer , Theodor Rüber , Thomas Schultz

{"title":"Deep learning based tractography with TractSeg in patients with hemispherotomy: Evaluation and refinement","authors":"Johannes Gruen , Tobias Bauer , Theodor Rüber , Thomas Schultz","doi":"10.1016/j.nicl.2025.103738","DOIUrl":null,"url":null,"abstract":"<div><div>Deep learning-based tractography implicitly learns anatomical prior knowledge that is required to resolve ambiguities inherent in traditional streamline tractography. TractSeg is a particularly widely used example of such an approach. Even though it has exclusively been trained on healthy subjects, a certain level of generalization to different pathologies has been demonstrated, and TractSeg is now increasingly used for clinical cases. We explore the limits of TractSeg by evaluating it on a unique dataset of 25 patients with epilepsy who underwent hemispherotomy, a type of surgery in which the two hemispheres are surgically separated. We compare results to those on 25 healthy controls who have been imaged with the same setup.</div><div>We find that TractSeg generalizes remarkably well, given the severity of the abnormalities. However, to our knowledge, we are the first to document cases in which TractSeg erroneously reconstructs (“hallucinates”) tracts that are known to have been surgically disconnected, and we found cases in which it implausibly continues tracts through obvious lesions. At the same time, TractSeg failed to reconstruct or undersegmented some tracts that are known to be preserved.</div><div>We subsequently propose a refinement of TractSeg which aims to improve its applicability to data with pathologies, by using its Tract Orientation Maps as an anatomical prior in low-rank tensor approximation based tractography such that tracking is guaranteed to continue only where presence of the tract is directly supported by the data (“data fidelity”). We demonstrate that our extension not only eliminates hallucinated tracts and reconstructions within lesions, but that it also increases the ability to reconstruct the preserved tracts, and leads to more complete reconstructions even in healthy controls. Despite these advances, we recommend caution and manual quality control when applying deep learning based tractography to patient data.</div></div>","PeriodicalId":54359,"journal":{"name":"Neuroimage-Clinical","volume":"45 ","pages":"Article 103738"},"PeriodicalIF":3.4000,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Neuroimage-Clinical","FirstCategoryId":"3","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2213158225000087","RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"NEUROIMAGING","Score":null,"Total":0}

引用次数: 0

Abstract

Deep learning-based tractography implicitly learns anatomical prior knowledge that is required to resolve ambiguities inherent in traditional streamline tractography. TractSeg is a particularly widely used example of such an approach. Even though it has exclusively been trained on healthy subjects, a certain level of generalization to different pathologies has been demonstrated, and TractSeg is now increasingly used for clinical cases. We explore the limits of TractSeg by evaluating it on a unique dataset of 25 patients with epilepsy who underwent hemispherotomy, a type of surgery in which the two hemispheres are surgically separated. We compare results to those on 25 healthy controls who have been imaged with the same setup.

We find that TractSeg generalizes remarkably well, given the severity of the abnormalities. However, to our knowledge, we are the first to document cases in which TractSeg erroneously reconstructs (“hallucinates”) tracts that are known to have been surgically disconnected, and we found cases in which it implausibly continues tracts through obvious lesions. At the same time, TractSeg failed to reconstruct or undersegmented some tracts that are known to be preserved.

We subsequently propose a refinement of TractSeg which aims to improve its applicability to data with pathologies, by using its Tract Orientation Maps as an anatomical prior in low-rank tensor approximation based tractography such that tracking is guaranteed to continue only where presence of the tract is directly supported by the data (“data fidelity”). We demonstrate that our extension not only eliminates hallucinated tracts and reconstructions within lesions, but that it also increases the ability to reconstruct the preserved tracts, and leads to more complete reconstructions even in healthy controls. Despite these advances, we recommend caution and manual quality control when applying deep learning based tractography to patient data.

查看原文本刊更多论文

求助全文

约1分钟内获得全文求助全文

来源期刊

Neuroimage-Clinical NEUROIMAGING-

CiteScore

7.50

自引率

4.80%

发文量

368

审稿时长

52 days

期刊介绍： NeuroImage: Clinical, a journal of diseases, disorders and syndromes involving the Nervous System, provides a vehicle for communicating important advances in the study of abnormal structure-function relationships of the human nervous system based on imaging. The focus of NeuroImage: Clinical is on defining changes to the brain associated with primary neurologic and psychiatric diseases and disorders of the nervous system as well as behavioral syndromes and developmental conditions. The main criterion for judging papers is the extent of scientific advancement in the understanding of the pathophysiologic mechanisms of diseases and disorders, in identification of functional models that link clinical signs and symptoms with brain function and in the creation of image based tools applicable to a broad range of clinical needs including diagnosis, monitoring and tracking of illness, predicting therapeutic response and development of new treatments. Papers dealing with structure and function in animal models will also be considered if they reveal mechanisms that can be readily translated to human conditions.