Explore Our Research
Temporal Multi-Modal Pipeline for Chronic Nonbacterial Osteomyelitis (CNO) Diagnosis Assistant and Early Treatment Response Prediction
Chronic nonbacterial osteomyelitis (CNO) demands precise lesion detection and proactive therapy monitoring. Our pipeline begins by applying a 3D U-Net combined with a Mask R-CNN head to whole-body STIR MRI scans, automatically generating pixel-level lesion masks and standardized CROMRIS scores, with clinician-in-the-loop refinements that dramatically reduce annotation time. These segmented features are then fused with lab values, demographic information, medication history and CROMRIS scores at each treatment stage within a temporal encoder. As new imaging and clinical data arrive, the model continuously refines its response predictions — mapping them into effective, partial or non-response categories — so that treatment plans can be adjusted in near real-time. In retrospective CNO cohorts, this approach can achieve expert-level segmentation accuracy and strong early-stage predictive performance, improving further as treatment progresses. By streamlining diagnosis, quantifying lesion burden and enabling personalized management, this framework has the potential to transform care for CNO patients as a tool for physician investigation.
CHOIR is a prospective observational study that seeks to understand CNO via collecting data from medical charts, patient surveys about their lived experiences. The main aims of CHOIR are to determine the effectiveness of commonly used medications in standard practice for patients with CNO, the long-term outcomes of CNO and the quality of life in affected patients. CHOIR is an international network including 20+ sites across the United States, Canada, Africa, Europe and Australia. Our goal is to create a unified team of researchers, healthcare providers and families to advance our knowledge in CNO. As of 2025, more than 600 patients across the world have been enrolled and we plan to enroll at least 2,000 patients and follow them for at least 10 years.
Our group led the international collaborative group following rigorous methodology of consensus-based and data-driven approach. To ensure the findings had broad relevance, we collected data from 31 centers across six continents. The international collaboration had four key phases. First, candidate items were proposed based on a survey of pediatric rheumatologists. Following this, criteria definition and reduction was achieved, and the group then performed criteria weighting using multicriteria decision analysis. Finally, weights and threshold scores were refined in a development cohort of 441 patients, and validated in a separate group of 514 patients. The new criteria have two initial steps. First, for a classification of CNO, children must be under the age of 18; have bone pain or musculoskeletal limitations that have been going on for at least six weeks; and have abnormal findings on X-ray or advanced imaging such as MRI. Second, they must not have any evidence of malignancy, infection, vitamin C deficiency or hypophosphatasia. If these two initial steps are satisfied, a score is then calculated across nine domains – five of which are clinical, the remaining four from pathology and laboratory measures – with a total score of at least 55 required for classification as CNO. The selected items are easy to apply in children without the absolute requirement for a bone biopsy. When tested in the validation cohort, the new criteria had a sensitivity of 82% and specificity of 98%. An online calculator has been developed by our team for public use.
- Funding source: European Alliance of Associations for Rheumatology (EULAR) and the American College of Rheumatology (ACR)
- Project lead: Haodong (Daniel) Chen, PhD
Dr. Farzana Nuruzzaman led this project in collaboration with our team and CARE Arthritis Ltd. Chronic nonbacterial osteomyelitis (CNO) is an autoinflammatory bone disease that causes pain and destruction. Appropriate diagnosis and treatment of CNO is needed to prevent these complications. Radiological imaging aids in the assessment of disease activity in CNO and guides physicians about medication recommendations. However, radiological reports are often descriptive and prone to interrater discrepancies. Our team has created a reliable scoring system for magnetic resonance imaging and has the potential as research measuring tool in clinical trials. We have refined this tool by adapting this paper-based scoring tool into a web-based platform. Thus, radiological interpretation of imaging in CNO can be standardized and accessible for radiologists at multiple sites of future clinical trials.
- Funding source: Arthritis Foundation, CARRA, Kaila’s Komfort
Longitudinal Course and Biopsychosocial Risk Factors for Chronic Pain in Children With Chronic Nonbacterial Osteomyelitis and Chronic Recurrent Multifocal Osteomyelitis: A Prospective Study
The team hypothesizes that there will be differences in patients' pain trajectories over the study period. We would like to analyze the differences between pain intensity, interference and behaviors trajectories with relation to CDAS. We would like to examine baseline biopsychosocial factors including sex; age; concurrent disease/comorbidities; fatigue; anxiety and depressive symptoms; socioeconomic status; and neighborhood distress as correlates of the trajectories for each pain outcome. Additionally, the impact of length of illness, medication class, inflammatory markers and other PROMIS measure discrete data elements on patient pain trajectories.
- Project lead: Natalie L. Rosenwasser, MD, co-director, CRMO Program
After first-line treatment of CNO patients with NSAIDs, it is not known what is the best second-line agent. There have been limited studies with mixed results and low patient numbers that have not yet answered this question. Using the power of the CHOIR registry (Chronic Nonbacterial Osteomyelitis International Registry), and the validated disease activity score, we now have both the numbers and validated tools to assess response to treatment. This project will compare the efficacy of TNF inhibitors with that of bisphosphonates. Next steps include identifying baseline characteristics of patients who respond to each class of medication as a first step in determining a more personalized algorithm for treatment of pediatric CNO.
- Project lead: Natasha Moussouras, MD, PhD, clinical fellow
Dr. Melissa S. Oliver led this project in collaboration with OMERACT TAG and CNO workgroup. OMERACT is an international research group committed to improving outcomes for patients with rheumatic diseases. OMERACT supports the development of outcome measurements for use in clinical trials. The CNO and SAPHO working group was established in 2019 with the goal of developing a core domain set and core outcome measurement sets for adults and children with CNO and SAPHO. Different criteria have been reported to define clinical responses to treatments in CNO, but a core domain set (CDS) recommendation for CNO has never specifically been made. The CNO OMERACT working group is currently working on the development of our CDS. Our group has completed a scoping review, virtual online focus groups and an online discussion board. We have generated a comprehensive list of candidate core domains. This list will be narrowed down through a series of Delphi surveys, and voting will take place in the final core domain set.
Whole Blood Transcriptomic Analysis of PBMCs From Patients With CNO
A preliminary run of transcriptome analysis of whole blood RNA (four healthy controls, three baseline samples from CNO patients and five follow-up samples from CNO patients after treatments) using NanoString nCounter showed gene expression differentially displayed in whole blood of children with active CNO versus that of healthy children and the alteration of pathways by treatment in children with CNO. IL23A was among genes overexpressed in children with active CNO comparing to healthy children. The IL-6 pathway was upregulated after effective treatment in children with CNO. The next step is to expand our sample size and examine the intraindividual and interindividual variation and reproducibility. Furthermore, the association between candidate markers and disease status will also be investigated.
- Collaborator: Polly Ferguson, MD
CNO can lead to chronic pain, growth disturbances and other long-term complications in children. It is often associated with comorbid conditions such as inflammatory arthritis and psoriasis, but much remains unknown about the prevalence of these risk factors on a global scale long-term. The CHronic nonbacterial Osteomyelitis International Registry (CHOIR) is a multinational research initiative designed to better understand the disease outcomes, medication usage and prevalence of comorbidities in children with CNO. The registry includes children with CNO from 17 global sites who were enrolled between August 1, 2018, and December 31, 2024. Through CHOIR, we were able to collect detailed clinical data at diagnosis and during follow-up visits over five years. This includes information on disease activity, imaging findings, growth measurements, medication use and the occurrence of any comorbidities or events of interest. By analyzing this longitudinal data, the project aimed to estimate the prevalence and progression of long-term risks and complications in CNO and identify patterns that can inform future care. This global effort represents one of the most comprehensive studies to date focused on the long-term outcomes of children living with CNO.
- Project lead: Jessica Kent, MBA, medical student, Medical College of Wisconsin
Osteoclasts are specialized cells responsible for bone resorption. Their overactivation can contribute to bone damage and structural changes as seen in conditions such as Paget’s disease, bone tumors and osteoporosis. This study investigated whether osteoclast formation (osteoclastogenesis) and serum levels of two key bone regulators, receptor activator of nuclear factor kappa-B ligand (RANKL) and osteoprotegerin (OPG), are overactivated in children with CNO. Additionally, the study assessed the effect of abatacept, a CTLA-4 analog, on osteoclast development. Children with CNO, inactive juvenile idiopathic arthritis (JIA) and healthy controls were enrolled at Seattle Children’s Hospital. PBMCs were cultured with varying serum concentrations and with and without abatacept to assess osteoclast counts across patient groups. Additionally, serum levels of RANKL and OPG were measured across all patient groups. This research aims to advance our understanding of pathogenesis of bone remodeling in patients with CNO and identify potential novel therapies in CNO including abatacept.
- Funding source: Bristol-Myers Squibb
- Collaborator: James E. Cassat, MD, PhD
This study developed a new scoring system to help doctors decide when a bone biopsy is necessary for children who may have chronic nonbacterial osteomyelitis (CNO). Because CNO can look like other conditions, a biopsy is often required to confirm the diagnosis. The research team used data from nearly 1,000 international patient cases to create a simple scoring system based on a patient's imaging and routine clinical information. The score places patients into low, mid and high categories, which helps guide the decision-making process. The study found that most CNO cases fell into the high-score category, while a majority of the "mimic" conditions (those that looked like CNO but were not) were in the low-score category. This new tool can help doctors more confidently determine who is a strong candidate for CNO and when to consider alternative diagnoses, potentially reducing the need for an invasive bone biopsy in many cases.
- Collaborators: Rajdeep Kaur Pooni, MD; Kevin Cain; and Sarah Menashe, MD
A Delphi survey on priorities for defining MDA and flare was sent to CNO patient/caregivers and CNO workgroup. Based on responses, key parameters from our multisite prospective registry (CHronic nonbacterial Osteomyelitis International Registry, CHOIR) were extracted to create three cohorts to define MDA, flare and MCID. Missing variables were imputed. A global expert panel of 13 pediatric rheumatologists and seven patient/caregivers voted on whether a clinical scenario was considered MDA (a single time point assessment), flare (worsening disease compared to prior visit) or MCID (meaningful improvement compared to prior visit). Consensus was defined as ≥80% agreement using nominal group technique. A decision tree or ROC analysis were used to determine the cutoff points. The model was validated using 10-fold cross-validation to ensure generalizability.
- Project leads: Farzana Nuruzzaman, MD, and Natalie L, Rosenwasser, MD
- Funding sources: Arthritis Foundation and CARRA bridge funding 2024, Kalia's Komfort
This study investigated whether doctors can accurately assess a patient's joint inflammation during a virtual visit using tools like a video exam and thermal imaging. Researchers had a group of virtual doctors review patient information, followed by a video of a joint exam and then thermal images. They then compared these assessments to what an in-person doctor found. The results showed that while virtual doctors generally agreed with in-person doctors on a patient's overall disease status, they were not very good at identifying specific inflamed joints. Interestingly, adding more information like the video and thermal images did not consistently improve the accuracy, and thermal imaging in particular did not provide any added benefit. The study concluded that a combination of patient history and a video exam can give doctors a moderately consistent assessment of joint health during telemedicine appointments.
- Funding source: Rheumatology Research Foundation
- Project lead: Ian Muse, BS
- Collaborators: Jenna L. Thomason, MD, MPH; Rajdeep Kaur Pooni, MD; Tzielan Lee, MD; Grant C. Hughes, MD; and Elizabeth Ferucci, MD, MPH
Development and Validation of a Novel Algorithm of Thermal Imaging to Detect Inflammatory Knee Arthritis; Feasibility of Implementation in Home Setting
The first study aimed to improve the accuracy of thermal imaging for detecting arthritis in children. We created a new measurement called TAWiC (Temperature After within-limb Calibration), which compares the temperature of a joint to a nearby body part, like the mid-tibia. The study found that TAWiC was much better than just using the absolute temperature of the joint for identifying inflamed areas. When we tested the method on a new group of children, the TAWiC score was highly specific, meaning it was very good at correctly identifying joints without arthritis. The study concluded that TAWiC is a promising and reliable tool for using infrared thermal imaging to detect arthritis in the knees and ankles of children.
The second study investigated the use of thermal cameras to detect arthritis in children. We first found that physical activity can change a person's body temperature, which needs to be accounted for when taking a reading. We then tested whether a smartphone-attached thermal camera could work as well as a professional handheld camera. The results were very promising: the smartphone camera performed similarly to the professional one, showing high sensitivity and specificity in detecting joint inflammation. The study concluded that this type of accessible technology is a valid and useful tool for identifying arthritis.
The third study investigated the feasibility of using a smartphone-attached camera for families to take thermal images of their child's joints at home to help detect arthritis. Our team compared these at-home images to those taken in a clinic and found that while the absolute temperatures were different, a special calibrated measurement called TAWiC (Temperature After within-limb Calibration) showed a good correlation between the two settings. The study demonstrated that the at-home imaging was just as effective as in-clinic imaging at detecting arthritis, with similar sensitivity and specificity. Using a smartphone-based thermal camera operated by families at home is a feasible method for arthritis detection and holds promise for future use in telemedicine.
Read Feasibility of applying infrared thermal imaging for home monitoring of arthritis in children.
- Funding sources: Seattle Children's Research Institute, Washington Research Foundation
- Project lead: Joshua Scheck
- Collaborators: Savannah Partridge, PhD; Stephen Chee-Yung Wong, MD; and Erin Balay-Dustrude, MD, MS
Development of Software as a Medical Device (SaMD) by Using Infrared Thermal Camera to Detect Arthritis in Children
A new software prototype, called the "Joint Thermometer," has been developed to simplify and automate the process of detecting arthritis using a thermal camera. Designed to be installed on a tablet or smartphone, the software can quickly analyze images from a connected camera in under 30 seconds. The goal is to provide a user-friendly and reliable tool that can produce a report for a doctor's review or for a patient's long-term monitoring, making this type of technology more accessible and practical for clinical use. Our innovation team assisted the development of job aids for each party and followed the greenlight guru process for optimization.
- Funding source: CTIP
- Project leads: Josh Scheck, Weston Cox, Nagu Rangan
- Collaborators: Promenade Inc., Chuck Magness, Lyra Hancock, Andrea Obradors-Puerto
Comparative effectiveness study of current treatments in children with Chronic Nonbacterial Osteomyelitis (CNO) NIH-NIAMS (1R56AR083929-01)
This research project aims to determine the most effective length of treatment for children with chronic nonbacterial osteomyelitis (CNO) who respond well to NSAIDs. The study will track patients at a single site, comparing those who take NSAIDs for a shorter period (three to six months) to those who take them for a prolonged period (12 to 24 months). The main goal is to see if patients on the shorter course of treatment are more likely to have their symptoms return (a "flare") after stopping the medication. The research will also look at whether the disease activity score changes differently between the two groups. In addition to these primary goals, the project will work to get a single IRB approval across all U.S. sites, identify challenges in enrolling patients and monitor for side effects and patient treatment preferences.
The Pathogenesis of Chronic Nonbacterial Osteomyelitis
The main objective of using digital spatial profiling on a bone sample is to create a highly detailed map of the tissue. Think of it like a GPS for your cells, but instead of showing roads and buildings, it shows exactly where different types of cells, genes and proteins are located. For an autoimmune bone disease, this is incredibly powerful. An autoimmune disease is when the body's immune system mistakenly attacks its own tissues. Instead of just grinding up the entire bone sample and seeing a mixed-up list of what's inside, digital spatial profiling allows researchers to:
- Pinpoint the attack: Identify the precise spots where immune cells are gathering and infiltrating the bone tissue.
- Listen to the conversation: See exactly which genes and proteins are active in those specific areas. This reveals the "message" the immune cells are sending that tells the body to attack the bone.
- Find new treatment targets: By mapping out the specific cellular pathways and molecules involved in the attack, researchers can identify new targets for medication that could stop the immune response without affecting the entire body.
- Funding source: Kaila’s Komfort
- Collaborator: James E. Cassat, MD, PhD
Phenocopy of CNO in Humanized Animal Models
It remains unknown whether the circulating immune cells or residential cells within the bone sites are major contributors to the inflammation within patients with CNO. One possible approach to identify the exact causal cellular or humoral component of dysfunctional immune system if any is through the transferring cells from patients to the host animal models and examine the phenotype as well as the molecular signatures of the affected bone sites in animal models. A similar disease, psoriatic arthritis, was successfully phenocopied in these animal models.
- Collaborator: Maria de la Luz Garcia-Hernandez, PhD