Recent Prospects for the Diagnosis and Management of Multiple Sclerosis: A Succinct Review

1. INTRODUCTION

Multiple sclerosis (MS) is a chronic inflammatory autoimmune disorder resulting from the interplay of genetic, environmental, and epigenetic factors [Figure 1], characterised by irreversible demyelination of the myelin sheath surrounding neurons in the brain and spinal cord, leading to impaired communication between the brain and the rest of the body.^[1] The diagnosis and treatment of MS are challenging because of the heterogeneous nature of the disease in terms of onset, progression, and response to therapy.^[2] In patients with MS, disruption of the blood–brain barrier (BBB) enhances immune cell infiltration into the central nervous system (CNS) parenchyma, resulting in the formation of lesions that are considered early hallmarks of the disease. However, these lesions are not always easily detected using traditional diagnostic methods.^[3]

Close

Figure 1: The interplay of genetic, environmental, and epigenetic factors in MS. Reproduced with permission from Boutitah-Benyaich et al. EBV: Epstein–Barr virus, HLA: Human leukocyte antigen, MS: Multiple sclerosis.

Export to PPT

Current diagnostic strategies primarily rely on clinical manifestations, magnetic resonance imaging (MRI), and cerebrospinal fluid (CSF) analysis. However, these approaches face limitations due to the potential for misdiagnosis in up to 20% of cases, largely because of the lack of highly sensitive and specific biomarkers for MS. Moreover, distinguishing MS from other neurological disorders, such as neoplasms and autoimmune encephalitis, remains challenging.^[4] The unpredictable nature of the disease, characterised by alternating periods of relapse and remission, together with the difficulty of delivering therapeutics across the BBB, further limits the effectiveness of current treatment strategies.^[3] Current treatment approaches for MS include pharmacotherapy [Figure 2], gene therapy, cell-based therapy, and monoclonal antibody therapy.^[5] However, these therapeutic options are limited by the absence of a definitive cure, the potential for adverse drug reactions, and pharmacokinetic challenges such as poor solubility, limited bioavailability, and poor tolerability.^[6] These limitations have driven the development of more advanced therapeutic approaches, such as antigen-specific immunotherapies [Figure 3].

Close

Figure 2: Different disease-modifying therapies for MS. Reproduced with permission from Yang et al. doi: 10.3389/fneur.2022.824926, under creative commons license 4.0. BBB: Blood brain barrier, CNS: Central nervous system, AHSCT: Autologous haematopoietic stem cell transplantation.

Export to PPT

Close

Figure 3: Antigen-specific immunotherapies for MS, including a) Peptide and protein-based approaches, b) DNA vaccination approach, c) Biological carrier approaches, d) Synthetic carrier approaches, and e) Cellular immunotherapy approaches. Reproduced with permission from Boutitah-Benyaich et al. APC: Antigen-presenting cells, NP: Nanoparticle, MS: Multiple sclerosis, CNS: Central nervous system, BBB: Blood brain barrier, PBMC: Peripheral blood mononuclear cells, PEG: Polyethylene glycol, PLGA: Polylactic-co-glycolic acid, MHC: Major histocompatibility complex, APL: Altered peptide ligands, GM-CSF: Granulocyte–macrophage colony-stimulating factor, IL: Interleukin, EV: Extracellular vesicle.

Export to PPT

Nanoparticle (NP)-based drug delivery systems for neurodegenerative diseases represent a promising strategy for enhancing drug delivery to the CNS and overcoming challenges associated with the BBB. Theranostic nanocarriers are an emerging approach that combines both diagnostic and therapeutic functions, making them highly effective tools for the real-time management of MS. Their advantages arise from the ability of nanocarriers to deliver therapeutics to specific sites of inflammation within the CNS while simultaneously enabling real-time monitoring of disease progression through imaging or biosensing techniques For example, fluorescent dyes or magnetic nanoparticles can be utilised to track drug distribution and accumulation within the CNS, thereby facilitating the evaluation of therapeutic outcomes.^[7] However, identifying optimal and effective NP–drug combinations for targeted delivery remains challenging because of the vast number of possible NP and drug compound combinations. Artificial intelligence (AI) and machine learning (ML) algorithms have accelerated this process by predicting optimal drug and nanoparticle candidates.^[8] AI models can also simulate nanoparticle interactions with biological barriers and predict their pharmacokinetic behaviour within the body.^[9] In addition, AI holds significant potential for improving diagnostic accuracy by analysing complex datasets, including MRI scans, to classify disease stages and predict disease progression.

Therefore, the aim of this review was to discuss some prospects of AI-powered theranostic nanocarriers for the real-time monitoring and management of multiple MS. It succinctly highlights recent advances in nanotechnology, imaging modalities, and ML techniques for the design of functional nanoparticles, simulation of physiological conditions, and prediction of disease progression.

2. PATHOPHYSIOLOGY OF MS

3. DIAGNOSIS OF MS

The diagnosis of MS relies on the integration of clinical and paraclinical assessments to demonstrate dissemination in space (DIS) and dissemination in time (DIT), while excluding other conditions that may account for the observed symptoms.^[20]

4. POTENTIAL ROLE OF THERANOSTIC NANOCARRIERS IN MS

Theranostic nanocarriers represent a new era in the application of nanotechnology for medical intervention, enabling real-time monitoring of treatment progression. They are multifunctional nanosystems capable of simultaneously diagnosing diseases, delivering targeted therapies, and monitoring therapeutic outcomes. This multifunctional role makes theranostic nanocarriers particularly valuable in the management of complex and heterogeneous diseases such as cancer, neurodegenerative disorders, and other chronic conditions.^[32] Nanoparticles have been shown to increase drug half-life and enable controlled drug release. Although nanoparticles are generally considered safe for short-term use, further studies are required to fully evaluate their long-term safety profiles. Factors such as particle size, concentration, and target tissue significantly influence nanoparticle biocompatibility.^[7] Nanomedicine also offers enhanced stability for therapeutic molecules and may facilitate transport across the BBB, thereby improving targeted delivery to demyelinated regions in MS.^[33]

The design of theranostic nanocarriers depends on careful material selection, functional integration, and advanced targeting mechanisms to optimise both therapeutic and diagnostic performance.^[34] These nanosystems are engineered to co-deliver therapeutic and diagnostic agents and may integrate multimodal imaging capabilities with stimuli-responsive drug release systems to enhance treatment specificity and therapeutic efficacy.^[35] Furthermore, targeting strategies such as ligand functionalisation and biomimetic surface coatings are employed to improve targeting accuracy while minimising systemic toxicity.^[36] The ability of theranostic nanosystems to penetrate the BBB is particularly important when targeting neurological disorders such as MS. Among the emerging delivery strategies, the nose-to-brain route has gained considerable attention for the administration of nanomedicines in neurodegenerative diseases. This approach enables direct drug transport to the brain through the olfactory and trigeminal pathways, thereby bypassing the BBB and reducing systemic side effects.^[37] The BBB is formed by tightly connected endothelial cells supported by pericytes and astrocytes, which collectively protect the brain from xenobiotics and circulating toxins. Tight junctions, with intercellular gaps of approximately 1.4 nm, together with active efflux transporters, restrict the passage of most large molecules and nearly 98% of small-molecule drugs into the CNS. In addition to particle size, molecular polarity also significantly influences CNS drug delivery, as many neurotherapeutic agents require high lipophilicity to effectively penetrate the BBB.^[38,39]

NPs can cross the BBB through different mechanisms, as illustrated in Figure 4. The receptor-mediated transcytosis allows endogenous macromolecules to enter the brain through interactions with specific endothelial receptors. The passive diffusion is the most direct route of transcellular transport, which allows small lipophilic molecules to diffuse across the endothelial membrane. In carrier-mediated transport, nutrient transporters facilitate the entry of specific molecules, such as glucose and amino acids, through specialised membrane carriers. Adsorptive-mediated transcytosis is driven by electrostatic interactions between positively charged molecules and the negatively charged endothelial surface.^[39,40] The physicochemical properties of NPs, including small particle size, surface charge, hydrophobicity, and ligand-mediated targeting capabilities, influence their ability to cross the BBB.^[41,42]

Close

Figure 4: Mechanisms of NPs penetration of BBB. Reproduced with permission from Hersh et al. 2022 doi: 10.3390/ijms23084153, under creative commons licensing 4.0.

Export to PPT

Lipid-based nanocarriers can enhance BBB penetration efficiency owing to their lipophilic properties, which facilitate interactions with cellular membranes and promote endocytic uptake by brain endothelial cells.^[43] Liposomes and polymeric nanoparticles exhibit high biocompatibility and enable controlled or modified drug release, whereas dendrimers possess a well-defined molecular architecture that supports multifunctional targeting applications, and may facilitate receptor-mediated or adsorptive transcytosis following surface functionalisation.^[44,45] Other nanoparticles such as quantum dots provide superior optical imaging properties, particularly high resolution fluorescence with nanoscale precision, allowing real-time imaging to follow up the treatment and disease progression.^[46] In addition, magnetic nanoparticles can be externally guided using magnetic fields and also possess imaging capabilities, making them suitable theranostic agents. These nanoparticles can act as sensitive T2-weighted MRI contrast agents that are easily taken by the circulating macrophages. In MS, they can allow the detection and visualisation of active neuroinflammation and BBB breakdown by tracking their entry into the CNS. In addition, they can be used to deliver anti-inflammatory medications, such as glucocorticoids or interferon-beta, allowing for localised therapy at the site of lesion formation while monitoring the reduction in inflammation via MRI.

Studies have demonstrated that nanoparticles can enhance imaging contrast and improve the detection of demyelinating lesions within the CNS. However, several limitations, including biocompatibility concerns, potential toxicity, technical complexity, and limited imaging sensitivity, continue to challenge their application in real-time monitoring. Furthermore, the long-term safety profiles of many nanomaterials remain insufficiently characterised, limiting their broader clinical translation. Imaging modalities must also balance spatial resolution, tissue penetration depth, and quantitative accuracy to ensure effective diagnostic performance.

5. POTENTIAL ROLE OF AI IN MS

AI is transforming the field of nanomedicine by improving the design, development, and application of nanoscale materials for medical purposes. AI, refers to the ability of machines to perform cognitive functions such as learning, reasoning, and problem-solving. AI and ML have become valuable tools in nanomedicine, enabling researchers to design and optimise nanoparticle-based drug delivery systems. The integration of ML with perturbation theory (PT) has led to the development of perturbation theory–machine learning (PTML) models, which have demonstrated significant potential in predicting the biological activities of drug-decorated nanoparticles (DDNPs) without the need for extensive experimental trials.^[47]

Moreover, nanoparticle size and shape are critical determinants of drug delivery efficiency, influencing therapeutic efficacy and the occurrence of side effects. Large ML-based databases can be used to analyse the physicochemical properties and biological responses of nanoparticles in order to predict optimal formulation conditions.^[48] In addition to optimising nanoparticle design, AI demonstrates considerable potential in broader neurological applications, including early diagnosis and disease monitoring. AI also plays an important role in personalised nanomedicine by analysing clinical and genetic data to predict patient-specific therapeutic outcomes. This approach enables the development of customised nanomaterials with enhanced therapeutic efficacy and reduced adverse effects.^[49]

ML and deep learning (DL) have become valuable tools in the early diagnosis, monitoring, and personalised treatment of neurological diseases. In ML, computers are capable of learning patterns from data without explicit programming, whereas DL, a subset of ML, utilises multilayered neural networks to enable software systems to learn and perform tasks through exposure to large volumes of data.^[50] AI-integrated devices also enable continuous, non-invasive, and patient-centred monitoring. For example, smartwatches and wearable sensors can quantify tremors, detect seizures, and distinguish between movement disorders. In addition, smartphone applications and medication adherence tools provide tracking systems, digital performance assessments, and support long-term disease management.^[51]

AI has demonstrated high sensitivity in lesion detection through the interpretation of complex imaging modalities such as computed tomography (CT), MRI, and positron emission tomography (PET), often without the need for contrast enhancement. Studies have shown that AI-assisted interpretation of low-dose MRI scans can achieve diagnostic quality and accuracy comparable to those of full-dose scans. These advancements may reduce patient exposure to radiation and potentially harmful contrast agents while simultaneously streamlining diagnostic procedures.^[52]

Despite the significant advantages of AI algorithms in medical image analysis and their ability to improve the speed and accuracy of neurological diagnoses, AI technologies may also introduce ethical challenges and inequalities in access to advanced neurological care. One of the major concerns associated with AI systems is the “black box” phenomenon, particularly in DL models. Although these models can generate highly accurate predictions, their underlying decision-making processes are often not fully interpretable by healthcare professionals, raising concerns regarding the transparency and reliability of AI-assisted diagnoses and therapeutic decisions. This lack of interpretability may hinder trust and limit the clinical adoption of AI technologies, as healthcare providers may be reluctant to rely on systems they do not fully understand.^[53]

AI provides a more accurate, efficient, and personalised approach to the early detection of MS, disease progression monitoring, and individualised treatment planning. ML and DL techniques are used to analyse extensive datasets, including MRI, genetic, and clinical data, thereby enhancing the precision of MS diagnosis and management.^[54] AI also demonstrates considerable potential in drug discovery, both through the identification of novel therapeutic agents and the repurposing of existing medications for MS treatment.^[55] AI-based chatbots, which have recently become widely used in healthcare, may also support MS care by providing patient education, basic symptom guidance, and triage support through natural language interactions. These systems can improve patient engagement and data accessibility while reducing physician workload; however, they should be used only as supportive tools that require clinical validation and professional oversight.^[56]

According to evidence from recent systematic reviews, AI models have achieved approximately 81% accuracy, 76.92% sensitivity, and 74% specificity in distinguishing patients with MS from healthy individuals and from patients with other demyelinating or neurological disorders.^[57] The integration of AI-driven models with clinical expertise is expected to improve the accuracy and clinical relevance of diagnostic and prognostic predictions compared with either approach alone. Nevertheless, the reliability and generalisability of these predictive models remain limited because larger and more diverse datasets are still required for robust validation. AI algorithms can also enhance the imaging analysis of MS by detecting subtle MRI changes that may be overlooked by clinicians, thereby facilitating earlier diagnosis.

One of the most valuable applications of AI in MS is automated diagnosis and lesion detection using medical imaging modalities such as MRI. AI models can identify MS lesions and distinguish them from healthy brain tissue more efficiently and consistently than conventional manual assessment methods, thereby reducing reporting time and improving diagnostic accuracy.^[58] This potential was demonstrated in a study conducted by Eshaghi et al., who employed an unsupervised AI algorithm known as Subtype and Stage Inference (SuStaIn) to analyse MRI data from approximately 9,000 patients with MS. The model identified three novel MRI-based subtypes of MS rather than relying solely on traditional clinical classifications: cortex-led, normal-appearing white matter (NAWM)-led, and lesion-led subtypes. Each subtype was associated with distinct pathological mechanisms and progression patterns. The lesion-led subtype demonstrated the fastest disability progression and the strongest treatment response, whereas the NAWM-led subtype exhibited intermediate progression, and the cortex-led subtype showed the slowest disease progression. The findings highlighted the potential of AI to redefine MS subtypes, improve prognostic accuracy, and support personalised imaging-guided management strategies.^[59]

In another observational study, Montolío et al. applied machine learning techniques to diagnose MS and predict long-term disability progression based on clinical data and retinal nerve fibre layer (RNFL) thickness measurements obtained through optical coherence tomography (OCT). RNFL thickness is considered a valuable biomarker in MS assessment. Compared with MRI, OCT is rapid, cost-effective, and non-invasive. In the predictive models, OCT measurements were combined with clinical and demographic data, while regression models and classifiers were trained using MRI-derived variables to predict disability progression. The study included 108 patients with MS and 104 healthy controls, of whom 82 patients completed a 10-year follow-up. The resulting models achieved an accuracy of 87.7%, sensitivity of 87.0%, specificity of 88.5%, and precision of 88.7%.^[60]

6. TECHNOLOGY DEVELOPMENT PERSPECTIVE

AI and ML are expected to play a central role in nanomedicine research over the coming decade. These technologies can accelerate the discovery and optimisation of nanomedicine platforms by improving the prediction of nanoparticle design, therapeutic targeting, and treatment outcomes. However, the lack of large, high-quality, and standardised datasets remains a major challenge. Harmonised data-sharing frameworks and standardised data collection practices will be essential for the development of reliable and generalisable AI models.^[61] Ethical and legal concerns, including data privacy, informed consent, and the need to preserve patient autonomy and justice, also represent significant challenges in the implementation of AI-driven healthcare systems. In addition, unequal access to AI-based technologies may further exacerbate existing socioeconomic disparities in healthcare delivery.^[62]

AI also demonstrates considerable future potential for personalised treatment planning and accurate prediction of MS progression. Nevertheless, limitations related to data quality, model interpretability, and insufficient clinical validation continue to hinder the full realisation of AI benefits in real-world clinical applications. Further clinical validation studies are necessary to ensure the reliable integration of AI technologies into healthcare systems and support effective patient management.

7. CONCLUSION

Although AI shows significant promise for the medical field, particularly in improving the precision and accuracy of disease diagnosis and detection as well as advancing personalised medicine, several challenges continue to limit its widespread clinical application. One of the major limitations is the lack of large, high-quality, and standardised datasets required for the development of reliable AI models. In addition, clinical confirmation and physician oversight remain essential for accurate diagnosis and therapeutic decision-making. In contrast, AI applications in drug discovery and nanomedicine optimisation represent highly effective supportive tools that can reduce experimental time, research costs, and laboratory workload. In summary, AI is expected to complement clinical and pharmaceutical research practices rather than replace human expertise.