Evaluation of a deep learning segmentation tool to help detect spinal cord lesions from combined T2 and STIR acquisitions in people with multiple sclerosis - Scorecard - MDSpire

Evaluation of a deep learning segmentation tool to help detect spinal cord lesions from combined T2 and STIR acquisitions in people with multiple sclerosis

  • By

  • Baptiste Lodé

  • Burhan Rashid Hussein

  • Cédric Meurée

  • Ricky Walsh

  • Malo Gaubert

  • Nicolas Lassalle

  • Guilhem Courbon

  • Agathe Martin

  • Jeanne Le Bars

  • Françoise Durand-Dubief

  • Bertrand Bourre

  • Adil Maarouf

  • Olivier Outteryck

  • Clément Mehier

  • Alexandre Poulin

  • Camille Cathelineau

  • Jeremy Hong

  • Guillaume Criton

  • Sophie Motillon-Alonso

  • Augustin Lecler

  • Frédérique Charbonneau

  • Loïc Duron

  • Alexandre Bani-Sadr

  • Céline Delpierre

  • Jean-Christophe Ferré

  • Gilles Edan

  • François Cotton

  • Romain Casey

  • Francesca Galassi

  • Benoit Combès

  • Anne Kerbrat

  • April 4, 2025

  • 0 min

Share

Clinical Scorecard: Assessment of a deep learning segmentation method for identifying spinal cord lesions using combined T2 and STIR imaging in individuals with multiple sclerosis

At a Glance

CategoryDetail
ConditionMultiple sclerosis with spinal cord lesions
Key MechanismsDeep learning segmentation of spinal cord lesions on combined T2-weighted and STIR MRI sequences
Target PopulationPeople with multiple sclerosis undergoing spinal cord MRI
Care SettingSpecialized MS centers and radiology/neuroimaging clinical settings

Key Highlights

  • Spinal cord lesions occur in up to 80% of people with MS and are critical for diagnosis and prognosis.
  • International guidelines recommend acquiring at least two spinal cord MRI sequences (T2 and STIR) for lesion detection.
  • A deep learning tool using combined T2 and STIR sequences was developed and evaluated to aid clinicians in detecting spinal cord MS lesions.

Guideline-Based Recommendations

Diagnosis

  • Perform spinal cord MRI systematically during initial MS assessment using at least two sequences (T2-weighted and STIR).
  • Use combined analysis of T2 and STIR sequences to improve lesion detection sensitivity.

Management

  • Incorporate deep learning segmentation tools as an aid to improve clinician sensitivity and precision in detecting spinal cord lesions.
  • Train clinicians with standardized protocols before using AI-assisted tools.

Monitoring & Follow-up

  • Use longitudinal spinal cord MRI data to support lesion identification and ground truth establishment.
  • Record interpretation time and inter-reader variability to assess tool impact.

Risks

  • Potential for decreased precision due to increased sensitivity with AI assistance.
  • Artifact risks in spinal cord MRI due to small tissue size and pulsation artifacts require careful image acquisition and interpretation.

Patient & Prescribing Data

People with multiple sclerosis undergoing spinal cord MRI with both T2 and STIR sequences

Deep learning segmentation tools can improve lesion detection sensitivity (up to 0.89) but may reduce precision (around 0.64); clinical use requires balancing these factors.

Clinical Best Practices

  • Use multi-sequence spinal cord MRI (T2 and STIR) for comprehensive lesion assessment in MS.
  • Apply deep learning segmentation tools as adjuncts to radiologist interpretation to enhance lesion detection.
  • Ensure clinicians receive standardized training on AI tools and annotation protocols.
  • Conduct multi-reader studies and use expert adjudication to establish reliable ground truth for lesion identification.
  • Monitor interpretation time and inter-reader variability when implementing new diagnostic aids.

References

Original Source(s)

Evaluation of a deep learning segmentation tool to help detect spinal cord lesions from combined T2 and STIR acquisitions in people with multiple sclerosis

  • by Baptiste Lodé, Burhan Rashid Hussein, Cédric Meurée, Ricky Walsh, Malo Gaubert, Nicolas Lassalle, Guilhem Courbon, Agathe Martin, Jeanne Le Bars, Françoise Durand-Dubief, Bertrand Bourre, Adil Maarouf, Olivier Outteryck, Clément Mehier, Alexandre Poulin, Camille Cathelineau, Jeremy Hong, Guillaume Criton, Sophie Motillon-Alonso, Augustin Lecler, Frédérique Charbonneau, Loïc Duron, Alexandre Bani-Sadr, Céline Delpierre, Jean-Christophe Ferré, Gilles Edan, François Cotton, Romain Casey, Francesca Galassi, Benoit Combès, Anne Kerbrat

  • April 4, 2025

Related Content

NPs Top US Jobs List Again

The nurse practitioner profession claims the No. 1 spot across three categories in the U.S. News & World Report 2026 Best Jobs rankings for the third consecutive year.

AI Scribes Lag Clinicians on Note Quality

A VHA study across 11 vendors finds AI-generated primary care notes score lower than clinician-written notes, with the largest deficits in thoroughness, organization, and usefulness

Editorial: Diffusion-weighted imaging: advances and implementations in neurology