Radiomics involves deep quantitative analysis of radiological images for structural and/or functional information. – It is a phenomic assessment of disease to understand lesion microstructure, microenvironment and molecular/cellular function. – In oncology, it helps us accurately classify, stratify and prognosticate tumors based on if, how and when they transform, infiltrate, involute or metastasize, – Utilizing radiomics in clinical trials is exploratory, and not an established end-point. – Integrating radiomics in an imaging-based clinical trials involves a streamlined workflow to handle large datasets, robust platforms to accommodate machine learning calculations, and seamless incorporation of derived insights into outcomes matrix.
With interest we read the recent article by Pinal-Fernandez and Mammen,1 which comments on the paper by Spielmann et al2 and to a lesser extent on the contribution by Mariampillai et al3 4 and raises concerns about the artificial intelligence (AI)-driven approach used to define subgroups of patients with idiopathic inflammatory myopathy (IIM).
To illustrate this, Pinal-Fernandez and Mammen constructed a library of 1000 observations and selected the four variables using a multivariate normal distribution, thus finding a similar clustering as in the original paper by Spielmann et al.2 We share some of the concerns about unsupervised learning techniques raised by Pinal-Fernandez and Mammen.1 In this letter, we would like to highlight several aspects related to AI-driven methodologies.
Machine learning (ML) is a subset of AI that enables a computer to make decisions based on the large dataset. When applied to clustering, it will always give an ‘optimal’ solution for the number of clusters ‘present’ in a dataset. However, it is up to the human user’s discretion to determine whether those clusters exist. An ML algorithm determines a number of clusters by separating the datasets into the subgroups through a process of optimising (1) separation between each cluster to its greatest and (2) ensuring that within a cluster, the distance to the cluster centre for each point is the smallest. Such an algorithm is essentially trying to identify a number of optimal clusters that allow each cluster to be distinct from the others. The goal is to have tight individual clusters that are very distinguishable from the others. In any dataset, the algorithms will present an optimal solution to those or similar criteria, but it does not always mean those clusters are truly significant or meaningful.
Visualising the clusters using dimensionality reduction techniques such as principal component analysis or t-distributed stochastic neighbour embedding is vital for this process, in addition to more quantitative methods such as comparing intracluster variation, intercluster variation and silhouette scoring. That is why researchers using ML should ideally be ‘bilingual’ and understand both the mathematics and algorithms, as well the science and clinical meaning behind the results.
To conclude, we emphasise that, no doubt, ML has the potential to improve the stratification of patients with IIM if certain concepts of data science are followed as also pointed out by a task force of the European League Against Rheumatism for big data and AI.5 ML relies on large, standardised and curated datasets that require large patient cohorts. Due to the rarity of IIM, larger patient cohorts (such as the MyoNet/EuroMyositis)6 are required to generate quality data. Once larger and curated datasets are available, the ML approach is a powerful alternative to human judgement and can improve future classification criteria for IIM.4 7 8Today, we argue for the use of ML alongside expert decision, thus relying on augmented judgement when making the final decision on patient stratification especially when building AI-based models. Augmented intelligence has the potential for improved patient stratification in IIM.
In this commentary we discuss the potential of advanced imaging, particularly Dynamic Contrast Enhanced (DCE) magnetic resonance imaging (MRI) for the objective assessment of disease progression in rheumatoid arthritis (RA). We emphasise the potential DCE-MRI in advancing the field and exploring new areas of research and development in RA. We believe that different grades of bone marrow edema (BME) and synovitis in RA can be examined and monitored in a more sensitive manner with DCE-MRI. Future treatments for RA will be significantly improved by enhanced imaging of BMEs and synovitis. DCE-MRI will also facilitate enhanced stratification and phenotyping of patients enrolled in clinical trials.
Copyright © Author(s) (or their employer(s)) 2015.
Trials. 2015 Apr;7(178)_suppl doi: 10.1186/s13063-015-0693-2
Trial registration: http://www.ClinicalTrials.gov identifier: NCT01656278 (5 July 2012)
Rheumatoid arthritis (RA) is a chronic, progressive joint disease, which frequently leads to irreversible joint deformity and severe functional impairment. Although patients are treated according to existing guidelines and reach clinical remission, erosive progression still occurs. This demonstrates that additional methods for prognostication and monitoring of the disease activity are needed. Bone marrow edema (BME) detected by magnetic resonance imaging (MRI) has proved to be an independent predictor of subsequent radiographic progression. Guiding the treatment based on the presence/absence of BME may therefore be clinically beneficial. We present the design of a randomized controlled trial (RCT) aiming to evaluate whether an MRI-guided treatment strategy compared to a conventional treatment strategy in anti-CCP-positive erosive RA is better to prevent progression of erosive joint damage and increase the remission rate in patients with low disease activity or clinical remission.
The study is a non-blinded, multicenter, 2-year RCT with a parallel group design. Two hundred anti-CCP-positive, erosive RA patients characterized by low disease activity or remission, no clinically swollen joints and treatment with synthetic disease-modifying antirheumatic drugs (DMARDs) will be included. Patients will be randomized to either a treatment strategy based on conventional laboratory and clinical examinations (control group) or a treatment strategy based on conventional laboratory and clinical examinations as well as MRI (intervention group). Treatment is intensified according to a predefined treatment algorithm in case of inflammation defined as a disease activity score (DAS28) >3.2 and at least one clinically swollen joint (control and intervention groups) and/or MRI-detected BME (intervention group only). The primary outcome measures are DAS28 remission (DAS28 < 2.6) and radiographic progression (Sharp/vdHeijde score).
The perspectives, strengths and weaknesses of this study are discussed.
Copyright © 2019 The Authors. Published by Elsevier Inc. All rights reserved.
Biochemical Pharmacology. 2019 Jul;37 doi: 10.1016/j.bcp.2019.02.037. Epub 2019 Mar 1.
The success of disease-modifying osteoarthritis drug (DMOAD) development is still elusive. While there have been successes in preclinical and early clinical studies, phase 3 clinical trials have failed so far and there is still no approved, widely available DMOAD on the market. The latest research suggests that, among other causes, poor trial outcomes might be explained by the fact that osteoarthritis (OA) is a heterogeneous disease with distinct phenotypes. OA trials might be more successful if they would address and target a specific phenotype. The increasing availability of advanced techniques to detect particular OA characteristics expands the possibilities to distinguish between such potential OA phenotypes. Magnetic resonance imaging is among the key imaging techniques to stratify and monitor patients with changes in bone, cartilage and inflammation. Biochemical markers have mainly used as secondary parameters and could further delineate phenotypes. Moreover, post-hoc analyses of trial data have suggested the existence of distinct pain phenotypes and their relevance in the design of clinical trials. Although ongoing work in the field supports the concept of OA heterogeneity, this has not yet resulted in more effective treatment options. This paper reviews the current knowledge about potential OA phenotypes and suggests that combining patient clinical data, quantitative imaging, biochemical markers and utilizing data-driven approaches in patient selection and efficacy assessment will allow for more successful development of effective DMOADs.
Copyright © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ.
Annals of the Rheumatic Diseases. 2019 Jun;78(6) doi: 10.1136/annrheumdis-2018-214539. Epub 2019 Mar 16.
To unravel the hierarchy of cellular/molecular pathways in the disease tissue of early, treatment-naïve rheumatoid arthritis (RA) patients and determine their relationship with clinical phenotypes and treatment response/outcomes longitudinally.
144 consecutive treatment-naïve early RA patients (<12 months symptoms duration) underwent ultrasound-guided synovial biopsy before and 6 months after disease-modifying antirheumatic drug (DMARD) initiation. Synovial biopsies were analysed for cellular (immunohistology) and molecular (NanoString) characteristics and results compared with clinical and imaging outcomes. Differential gene expression analysis and logistic regression were applied to define variables correlating with treatment response and predicting radiographic progression.
Cellular and molecular analyses of synovial tissue demonstrated for the first time in early RA the presence of three pathology groups: (1) lympho-myeloid dominated by the presence of B cells in addition to myeloid cells; (2) d iffuse-myeloid with myeloid lineage predominance but poor in B cells nd (3) pauci-immune characterised by scanty immune cells and prevalent stromal cells. Longitudinal correlation of molecular signatures demonstrated that elevation of myeloid- and lymphoid-associated gene expression strongly correlated with disease activity, acute phase reactants and DMARD response at 6 months. Furthermore, elevation of synovial lymphoid-associated genes correlated with autoantibody positivity and elevation of osteoclast-targeting genes predicting radiographic joint damage progression at 12 months. Patients with predominant pauci-immune pathology showed less severe disease activity and radiographic progression.
We demonstrate at disease presentation, prior to pathology modulation by therapy, the presence of specific cellular/molecular synovial signatures that delineate disease severity/progression and therapeutic response and may pave the way to more precise definition of RA taxonomy, therapeutic targeting and improved outcomes.