Artificial intelligence in the detection of multiple sclerosis plaques: Can it influence the treatment decision?
Authors
Affiliations (7)
Affiliations (7)
- Laboratoire d'imagerie Biomédicale Multimodale (BioMaps), Université Paris-Saclay, CEA, CNRS, Inserm, Service Hopsitalier Frédéric Joliot, Orsay, France; Faculty of Medicine, University of Novi Sad, Novi Sad, Serbia.
- Pixyl Research and Development Laboratory, Grenoble, 38000, France.
- Center for Intelligent Imaging, Department of Radiology & Biomedical Imaging, University of California, San Francisco (UCSF), San Francisco, CA, USA.
- Faculty of Medicine, University of Novi Sad, Novi Sad, Serbia; Oncology Institute of Vojvodina, Sremska Kamenica, Serbia.
- Department of Neurology, Poissy-Saint-Germain-en-Laye Hospital, Poissy, France; CRC SEP IDF Ouest, Poissy-Garches, France.
- Radiology Department, Centre Hospitalier Lyon-Sud, Hospices Civils de Lyon, 165 Chem. du Grand Revoyet, 69495 Oullins-Pierre-Bénite, France; CREATIS, INSERM U1044, CNRS UMR 5220, UCBL1, 43 Bd du 11 Novembre 1918, 69100 Villeurbanne, France.
- Laboratoire d'imagerie Biomédicale Multimodale (BioMaps), Université Paris-Saclay, CEA, CNRS, Inserm, Service Hopsitalier Frédéric Joliot, Orsay, France; Department of Radiology, APHP, Hôpitaux Raymond-Poincaré & Ambroise Paré, Paris, France. Electronic address: [email protected].
Abstract
To monitor multiple sclerosis (MS) progression, follow-up MRIs are used to detect new or enlarging lesions (ELs), typically through manual comparison - a time-consuming, error-prone process. This retrospective study used Pixyl.Neuro.MS®, an AI-based tool, to assess whether AI-assisted readings improve lesion detection and influence treatment decisions. The study was conducted retrospectively on MS patients. For the comparison of previous and new MRIs, the deep-learning-based software Pixyl.Neuro.MS® was used. This tool performs lesion segmentation and characterization according to their temporal evolution, facilitating the analysis of new lesions and ELs. Subsequently, a new AI-assisted radiological report was generated and compared with the conventional radiological report. By integrating the AI report with neurological assessments, its potential impact on treatment decisions, in contrast to those based solely on the standard radiological report, was evaluated. In this cohort of 83 MS patients (mean age 49 years, predominantly female), MRI analysis performed by radiologists with AI assistance significantly outperformed standard radiological interpretation. New lesions were identified in 30.1 % of patients using AI-assisted analysis, compared to 14.6 % with conventional reporting (p < 0.001). ELs were detected in 33.7 % of patients through AI-supported evaluation (p < 0.001), while none were identified with standard interpretation. On average, radiologists, aided by AI, identified more new lesions per patient (0.82 vs. 0.46) and achieved a higher true-positive lesion count. Importantly, integrating AI-assisted findings with clinical data led to treatment modification in 10.8 % of patients, underscoring the potential clinical impact of this approach. Artificial intelligence may play a key role in improving detection of new lesions and ELs in patients with MS. The use of Pixyl.Neuro.MS® enhanced radiological interpretation, yielding a more comprehensive assessment of MRI findings compared to conventional analysis. This improved diagnostic precision contributed to revised treatment decisions in a subset of patients.