JIA C. Mala*a, N. ter Haar, M. Vicecon6 WP5: JIA WP5 and WP10 - - PowerPoint PPT Presentation

JIA

• C. Mala*a, N. ter Haar, M. Vicecon6

WP5: JIA

WP5 and WP10

OBJECTIVES

Brussels, 12th May 2016

To develop a personalized image-based JOINT BIOMECHANICAL MODELING to explore how biomechanical changes may influence the spread of arthritis or structural damage progression.

POTENTIAL PREDICTORS

Brussels, 12th May 2016

Build a data repository containing (clinical, laboratory, imaging, immunological and microbiota characterization) that will be integrated to obtain a MULTIDIMENSIONAL PATIENT SPECIFIC MODEL of the disease course.

DATA HAS GONE THROUGH THE DCV TOOL AND ARE CURRENTLY BEING PROCESSED BY KDD (WP16)

Brussels, 12th May 2016

VARIABLE N (%) 2015 N (%) 2016 SELF ASSESMENT PLAN

Pa6ents enrolled 123 170 150-185 6 months FU 80 (65) 146 (86%) 12 months FU 33 (27) 101 (60%) 18 months FU 2 (2) 58 (34%) 24 months FU 0 29 (17%) Rou6ne lab 123 (100) 170 (100) Luminex blood 93 (76) 138 (81%) 130-150 Luminex synovial fluid 36 (29) 69 (41%) 35-50 Stools 83 (67) 120 (71%) 100-130 Ultrasound 112 (91) 157 (92%) 140-170

DATA COLLECTION PROGRESS

DATA COLLECTION MRI & GCA

• AutomaGc segmentaGon and modelling

started – see presentaGon WP10

• Ankle MRI scoring system

Brussels, 12th May 2016

MRI CGA Assessment

plan Baseline

14 12

26 24 25-40

6 months

12 12

21 21

12 months

7 6

10 10

24 months

3 N/A

3 N/A

DISTENSION OF THE JOINT RECESS DUE TO THE PRESENCE OF AN ENHANCING THICKENED SYNOVIUM.

Brussels, 12th May 2016

GRADE 1 GRADE 2 GRADE 3 Score 0 is normal; score 1-3 (mild, moderate, severe) are by thirds of the presumed max volume of the enhancing tissue in the synovial compartment.

SYNOVITIS

Tibio.peroneo.talar 0,78 Talo-navicular 0,72 Subtalar-anterior 0,46 Subtalar posterior 0,78 Calcaneo-cuboid 0,88 Cuneo-navicular 0,67 Tarso metatarsal medial 0,71 Tarso metatarsal lateral 0,87

ICC: 0.93 (0.81-0.98) Weighted kappa

Brussels, 12th May 2016

RELIABILITY

Brussels, 12th May 2016

Tendon sheath thickening and enhancement aZer i.v contrast injecGon.

K=0.72

Intra Class CorrelaGon Coefficient=0.97 (0.90-0.99)

K=0.86 K=0.9

TENOSYNOVITIS

A sharply bone lesion, with correct juxta-articular localization which is visible in 2 planes.

Bone erosion was scored from 0 to 10 by the volume of the erosion as a proportion of the “assessed bone volume” by 10% increments.

Brussels, 12th May 2016

Distal Gbial epiphysis 0,54 Distal fibular epiphysis 0,68 Talus 0,50 Calcaneus 1 Navicular bone 0,61 Cubical bone 1 Medial cuneiform 0,61 Weighted kappa

ICC=0.77 (0.57-0.94)

BONE EROSION

Distal Gbial epiphysis 1 Distal fibular epiphysis 0,91 talus 0,81 calcaneus 0,80 Navicular bone 0,47 Cubical bone 0,85 medialcuneiform 0,70 Intermediate cuneiform 0,76 Lateral cuneiform 0,4

ICC= 0.83 (0.52-0.96) Weighted kappa

Each bone was scored separately using a 0–3 scale based on the proportion of bone with BMO as follows: 0, no BMO; 1, 1%–33% of bone with BMO; 2, 34%– 66% of bone with BMO; and 3, 67%–100% of bone with BMO.

Brussels, 12th May 2016

BONE MARROW OEDEMA

Irregularities and/or interruption of the cartilage surface

0= no cartilage damage; 1= cartilage damage involving 1%–33% of the cartilage surface; 2= damage involving 34%–66% of the cartilage surface; 3= damage involving 67%–100% of the cartilage surface; 4= ankylosis.

ICC = 0.64

Brussels, 12th May 2016

CARTILAGE DAMAGE

Pa6ent Biomarkers Sex Age MRI synoviGs score MRI tenosynoviGs Height (cm) Weight (kg) BL BL BL BL M6 M12 M18 M24 BL M6 M12 M18 M24 IGG-RF F 13.5 10 Flexor and fibular tendons 163 164 164 163 N/A 53 64 62 63 N/A IGG-AP F 9.4 10 Flexor and extensor tendons 138 139 140 145 150 41 42 44 45 50 OPBG- MT M 12.1 3 Posterior Gbial tendon 149 154 152 N/A N/A 47 46 48 N/A N/A

Brussels, 12th May 2016

BIOMARKERS

Pa6ent Outcomes MRI damage progression CarGlage damage progression Treatment response (ACR70) InacGve disease (Wallace criteria) Last vs. first

• bservaGon

Last vs. first

• bservaGon

M6 M12 M18 M24 M6 M12 M18 M24 IGG-RF No No Yes Yes Yes N/A Yes No Yes N/A IGG-AP No Yes No Yes Yes Yes No No Yes Yes OPBG-MT No No Yes Yes N/A N/A Yes No N/A N/A

Brussels, 12th May 2016

MODELLING OUTCOMES

Cytokine analysis

• Baseline blood plasma: 138 individual paGents

– IGG 62, UMCU 23, OPBG 53 paGents

• Goal: find biomarkers/profiles that predict

disease outcome

– 78 analytes

• Progress:

– Samples are shipped from all centers – Luminex assay scheduled for 19th of May – Data analysis done in June-July

Brussels, 12th May 2016

Cytokine analysis: selected analytes

cytokines chemokines receptors & binding proteins MMPs & alarmins

• ther

IL-1 bèta CCL2/MCP-1 IL-1RA MMP-1 OPN IL-6 CCL3/MIP-1 alpha IL-18BPa MMP-3 SclerosGn/SOST IL-10 CCL4/MIP-1 bèta IL-1RI MMP-8 Dkk1 IL-12 p70 CCL8/MCP-2 IL-1RII MMP-9 LepGn IL-13 CCL17/TARC TNF-RI S100A8/MRP8 ResisGn IL-15 CCL18/PARC TNF-RII TIMP-1 GM-CSF IL-17 CCL22/MDC sCD19 HSP70/HSPA1A Amphiregulin IL-17F CCL23/MPIF sIL-2R/sCD25 VimenGn NGF IL-18 CCL25/TECK sCD27 VEGF IL-22 CCL27/C-TACK sIL-6R/sCD126 sICAM IL-23 p19 CXCL5/ENA-78 IL-7R alpha sVCAM IL-25/17E CXCL8/IL-8 sVEGF-R1/Flt-1 IL-27 CXCL9/MIG sCD14 TNF alpha CXCL10/IP-10 CD40L/CD154 IFN alpha CXCL13/BLC IFN gamma LIGHT TWEAK MIF CHI3L1/YKL-40 TSLP LAP/TGF-1 MIC-1/GFD15

Brussels, 12th May 2016

Cytokine analysis: future assays

• Cytokine assays for end 2016:

– Blood plasma of inacGve paGents (now: 68 pt)

• Biomarkers for predicGon which paGents will stay

inacGve vs which ones will flare

• Progress: sGll sampling inacGve paGents

– Synovial fluid plasma (now: 73 pt)

• Biomarkers for predicGon which paGents will get

inacGve disease, and will respond well to intra-arGcular injecGon

• Progress: baseline completed, sGll sampling paGents

with persistent acGvity or flare

Brussels, 12th May 2016

WP10: JIA modelling

WP5 and WP10

Technical goals

• Development of arGculated geometric models of JIA affected joints
• PaGent-specific biomechanical models and simulaGons
• MulG-dimensional modelling of the disease course
• Reuse/adapt developed modelling tools (VPHOP & NMS Physiome)
• ExtracGon of biomarkers (ideally automated)

– Incorporate anatomical and modelling biomarkers into MDP database

Brussels, 12th May 2016

OperaGonal goals – YR1

ü Agree and implement imaging and gait analysis data collecGon protocols ² Develop paGent-specific whole body musculoskeletal dynamics model ² Extend model to include detailed foot dynamics model Ø Start data collecGon – Quality assurance Ø Model few cases using manual processing for feasibility and validaGon Ø Automate data processing Ø Develop paGent-specific joint finite element model Ø Apply automated data processing to all paGents Ø Extract anatomical and funcGonal biomarkers Ø Generate full mulGscale paGent-specific models for all paGents Ø Extract biomechanics biomarkers

Brussels, 12th May 2016

OperaGonal goals – Y2

ü Agree and implement imaging and gait analysis data collecGon protocols ü Develop paGent-specific whole body musculoskeletal dynamics model ü Extend model to include detailed foot dynamics model ü Start data collecGon – Quality assurance ü Model few cases using manual processing for feasibility and validaGon ² Automate data processing ² Develop paGent-specific joint finite element model Ø Apply automated data processing to all paGents Ø Extract anatomical and funcGonal biomarkers Ø Generate full mulGscale paGent-specific models for all paGents Ø Extract biomechanics biomarkers

Brussels, 12th May 2016

Unexpected challenges in using MRI data

Brussels, 12th May 2016

• Successful automated segmentaGon
• f 42 bones, except pelvis and foot

at month 6.

• CarGlage layer at the ankle joint not

clearly visible in the images, hence not suitable for FE modelling

From FE to Hertz contact model

Brussels, 12th May 2016

Adopted countermeasures

Brussels, 12th May 2016

• Manual segmentaGon of feet and

pelvis bones for month-6 data.

• Reduced order model of

arGcular contact (Hertzian contact model) to reduce sensiGvity to quality of input data.

OperaGonal goals – YR3

ü Agree and implement imaging and gait analysis data collecGon protocols ü Develop paGent-specific whole body musculoskeletal dynamics model ü Extend model to include detailed foot dynamics model ü Start data collecGon – Quality assurance ü Model few cases using manual processing for feasibility and validaGon ü Automated data processing (specific datasets processed manually) ü Develop paGent-specific joint model using Heartzian contact ² Generate full mulGscale paGent-specific models for all paGents ² Apply automated data processing to all paGents ² Extract anatomical and funcGonal biomarkers ² Extract biomechanics biomarkers

Brussels, 12th May 2016

Year 3 main progresses

• Improve adaptaGon strategies for automaGc segmentaGon
• Increased number of training data sets
• Models extended by integraGng other structures
• Lower limb modelling procedures established and confirmed
• Biomarker extracGon started

Brussels, 12th May 2016

11 datasets for training 23 datasets for training

Ankle/Lower limbs geometrical models

• AutomaGc segmentaGon of anatomical structures in MRI

datasets (input for biomechanical modelling)

Brussels, 12th May 2016

DetecGon and quanGficaGon of inflammaGon in ankle joints (D10.3)

Anatomical Biomarker extracGon

Noise reducGon Find regions of interest mulGlevel thresholding Vesselness filter Volumetric representaGon

• f inflamed area

Brussels, 12th May 2016

Next steps for anatomical modelling

• For both models:

– Work on image intensity normalizaGon – Improve automaGc segmentaGon

• For lower limb models:

– Increase number of training data sets – Extend model by skin and main muscles

• For biomarker extracGon:

– Improve vessel detecGon – Add quanGficaGon

Brussels, 12th May 2016

Biomechanical modelling progress

• Robust method to merge 0-6-12 months data established.
• Allows to process incomplete datasets.

Generic model parGally personalized PaGent specific model

+

• r

Brussels, 12th May 2016

Biomechanical modelling results

• PaGent IGG-RF: data available at 0-6-12 months

Month 0 Month 6 Month 12 Height 163 163.6 163.7 Mass 53.2 63.8 62.0 Involvement Both sides None Both sides Model

Brussels, 12th May 2016

Brussels, 12th May 2016

SimulaGons SimulaGons

Biomechanical modelling results

Both sides involved (severe) Both sides involved (less severe than month-0) Inac6ve disease

• Ankle contact forces (%BW)

Brussels, 12th May 2016

Month 0 Month 6 Month 12

MRI damage progression CarGlage damage progression Treatment response (ACR70) InacGve disease (Wallace criteria) Last vs. first

• bservaGon

Last vs. first

• bservaGon

M6 M12 M18 M6 M12 M18 IGG-RF No No Yes Yes Yes Yes No Yes

NOTE: this paGent had pain in the ankles despite having being classified as a responder

Clinical use and validaGon

Clinical needs addressed (as per the requirements list in WP2/WP13) Descrip6on of the model Valida6on and tes6ng outcome Do biomechanical alteraGons correctly discriminate responsive and non- responsive paGents? Predicts responsive and non-responsive paGents Area under the ROC curve of the best classifier provided by a logisGc regression of all anatomical, funcGonal and biomechanical parameters Do biomechanical alteraGons affect structural damage progression? Predicts structural damage progression and locaGon Area under the ROC curve of the best classifier provided by a logisGc regression of all anatomical, funcGonal and biomechanical parameters Do biomechanical alteraGons correctly predict the locaGon

• f arthriGs in the lower limbs?

Predicts the locaGon of arthriGs in the lower limbs Area under the ROC curve of the best classifier provided by a logisGc regression of all anatomical, funcGonal and biomechanical parameters

Brussels, 12th May 2016

Clinical Outcomes

Pa6ent Outcomes MRI damage progression CarGlage damage progression Treatment response (ACR70) InacGve disease (Wallace criteria) Last vs. first

• bservaGon

Last vs. first

• bservaGon

M6 M12 M18 M24 M6 M12 M18 M24 IGG-RF No No Yes Yes Yes N/A Yes No Yes N/A IGG-AP No Yes No Yes Yes Yes No No Yes Yes OPBG-MT No No Yes Yes N/A N/A Yes No N/A N/A

Brussels, 12th May 2016

Clinical Biomarkers

Pa6ent Biomarkers Sex Age MRI synoviGs score MRI tenosynoviGs Height (cm) Weight (kg) BL BL BL BL M6 M12 M18 M24 BL M6 M12 M18 M24 IGG-RF F 13.5 10 Flexor and fibular tendons 163 164 164 163 N/A 53 64 62 63 N/A IGG-AP F 9.4 10 Flexor and extensor tendons 138 139 140 145 150 41 42 44 45 50 OPBG- MT M 12.1 3 Posterior Gbial tendon 149 154 152 N/A N/A 47 46 48 N/A N/A

Brussels, 12th May 2016

Anatomical Biomarkers

Pa6ent Biomarkers

Foot length [mm] Knee-ankle joint distance [mm] Heel pad thickness [mm] Inferior calcaneal inclina6on [degrees] Chopart’s angle [degrees] Talus volume [mm3] Calcaneus volume [mm3] Month 0/6/12 Month 0/6/12 Month 0/6/12 Month 0/6/12 Month 0/6/12 Month 0/6/12 Month 0/6/12

IGG-RF

R: 230/NA/260 L: 225/NA/262 R: NA/366/NA L: NA/364/NA R: 13/NA/13 L: 13/NA/15 R: 30/NA/28 L: 27/NA/26 R: 57/NA/62 L: 60/NA/60 R: 28222/NA/28395 L: 28395/NA/27912 R: 59437/NA/60656 L: 60175/NA/63006

OPBG-MT

R:235/NA/241 L: 240/NA/245 R: NA/342/NA L: NA/344/NA R: 16/NA/13 L: 18/NA/14 R:28/NA/22 L: 17/NA/21 R: 61/NA/59 L: 62/NA/57 R: 25458/NA/29217 L: 25736/NA/26361 R:43912/NA/55316 L:43987/NA/53612

IGG-AP

R: NA/NA/209 L: NA/NA/213 R: NA/291/NA L: NA/290/NA R:13/NA/12 L:16/NA/13 R:27/NA/26 L:24/NA/26 R:67/NA/64 L:64/NA/65 R: 20230/NA/21543 L: 19193/NA/22471 R:35979/NA/42844 L:29538/NA/42250

Brussels, 12th May 2016

Pa6ent

Biomarkers

Ankle contact forces [%BW] Max ant-post GRF [BW] Max ver6cal GRF [BW] Timing of max ver6cal GRF [% of stance] Maximum ankle joint moment [BWm] Timing of max ankle joint moment [% of stance] Max plantarflexion angle in stance [deg] Month 0/6/12 Month 0/6/12 Month 0/6/12 Month 0/6/12 Month 0/6/12 Month 0/6/12 Month 0/6/12

IGG-RF

R:419/517/466 L:450/499/481 R: 0.17/0.33/0.29 L: 0.17/0.34/0.28 R: 1.01/1.45/1.37 L: 1.05/1.45/1.36 R: 31/72/80 L: 41/74/80 R: 0.13/0.20/0.18 L: 0.13/0.20/0.19 R: 78/82/82 L: 76/83/82 R: 12/9/5 L: 16/11/10

OPBG- MT

IN PROGRESS R: 0.16/0.21/0.21 L: 0.15/0.17/0.19 R: 1.05/1.06/1.14 L: 1.04/1.04/1.14 R: 62/53/66 L: 44/62/43 R: 0.22/0.15/0.16 L: 0.14/0.20/0.25 R: 65/78/80 L: 67/63/53 R: 21/20/19 L: 19/18/20

IGG-AP

IN PROGRESS R: 0.18/0.14/0.19 L: 0.18/0.16/0.19 R: 1.03/1.02/1.07 L: 1.02/1.05/1.10 R: 80/58/67 L: 79/69/74 R: 0.12/0.12/0.13 L: 0.12/0.13/0.13 R: 82/81/81 L: 81/80/76 R: 23/24/23 L: 18/20/20

Biomechanical Biomarkers

Brussels, 12th May 2016

Deliverables

• D5.3: Report baseline and intermediate follow

up data (submised)

• D10.3: Report on biomarker extracGon

(submised)

• D10.4: Biomechanical simulaGon based on

image based modelling and gait analysis (submised)

Brussels, 12th May 2016

Future Work

• Processing all available datasets
• Extract anatomical, funcGonal and biomechanical

biomarkers for all paGents

• Explore which biomarkers, alone or in

combinaGon provide the best straGficaGon for disease course

• D10.5: Report on mulGdimensional modelling of

disease course (M42)

Brussels, 12th May 2016

JIA biomechanical models ValidaGon

Marco ViceconG

Models validaGon - 2015

• Solvers verificaGon

– PredicGons compared to analyGcal benchmarks

• Models verificaGon

– ODE: PreservaGon of momentum – OpGmiser: residuals, local minima – Solid elasGcity (FE): mesh convergence

• SensiGvity analysis

– Monte Carlo on selected parameters

• ValidaGon

– Based on draZ ASME V&V-40 (model credibility) – Model credibility levels (L1-L3) – Progressive increase of benchmark complexity

• ConfirmaGon

– Consistency, sensiGvity, specificity, discriminaGve power

Year 1&2

Brussels, 12th May 2016

Models validaGon - 2016

• Solvers verificaGon

– PredicGons compared to analyGcal benchmarks

• Models verificaGon

– ODE: PreservaGon of momentum – OpGmiser: residuals, local minima – Solid elasGcity (FE): mesh convergence

• SensiGvity analysis

– Monte Carlo on selected parameters

• ValidaGon

– Based on draZ ASME V&V-40 (model credibility) – Model credibility levels (L1-L3) – Progressive increase of benchmark complexity

• ConfirmaGon

– Consistency, sensiGvity, specificity, discriminaGve power

Brussels, 12th May 2016

ValidaGon: summary 2015

L1 = is a populaGon member L2 = close to the average L3 = close to the individual

Whole limb MSK Dynamics Foot MSK dynamics Ankle finite element model Code Verifica,on OK OK OK Verifica,on OK OK To be done Sensi,vity OK OK To be done Valida,on: theore,cal L2 L2 To be done Valida,on: in vitro NA NA To be done Valida,on: ex vivo NA NA To be done Valida,on: in vivo L1 To be done To be done Confirma,on, clinics Discriminant Consistent To be done

Brussels, 12th May 2016

ValidaGon: summary 2016

L1 = is a populaGon member L2 = close to the average L3 = close to the individual

Whole limb MSK Dynamics Foot MSK dynamics Ankle Hertzian model Code Verifica,on OK OK OK Verifica,on OK OK OK Sensi,vity OK OK OK Valida,on: theore,cal L2 L2 NA Valida,on: in vitro NA NA OK Valida,on: ex vivo NA NA L1 Valida,on: in vivo L3 L3 (knee) NA Confirma,on, clinics Discriminant Consistent On going

Brussels, 12th May 2016

Generic scaled model

• Vs. paGent-specific model

Prinold et al. (2016), Annals of Biomedical Engineering, 44, 247-257.

SensiGvity analysis: personalisaGon

MRI-based Scaling-based

Brussels, 12th May 2016

SensiGvity analysis: soZ Gssue artefact (STA) - Methods

Brussels, 12th May 2016

1. Synthesis of baseline marker trajectories during walking for three musculoskeletal models

STA Model 2 STA Model 1

2. STA modelling 3. SensiGvity analysis to assess:

• Joint Angles
• Joint Moments
• Muscle Forces
• Joint Contact

Forces

[Bonci et al. 2014] [Cheze et al. 1995, Dumas et al. 2009, Rozumalski et al. 2007, Tranberg et al. 1998]

Arnold Lower Limb (ALLM) Gait 2392 (G2392) London Lower Limb (LLLM)

SensiGvity analysis: soZ Gssue artefact (STA) – Results

Brussels, 12th May 2016

• 5th and 95th percenGles are in solid lines
• 50th percenGle is in dosed lines

ALLM G2392 LLLM Lamberto G., Martelli S., Cappozzo A., Mazzá C., Journal of Biomechanics, under review

Ankle dorsiflexion angle and moment Ankle contact force

ALLM G2392 LLLM

Repeatability analysis (D10.4)

• IdenGfied a repeatable set of 22

anatomical landmarks.

• It is criGcal to derive the model

parameters (joint axes) from the images instead of gait data.

• Some muscles asachments are

criGcal (triceps surae, Gbialis ant/post, peroneus long)

Brussels, 12th May 2016

Brussels – 11-12 May 2016

• Muscle acGvaGons esGmated by the model are

compared against experimental EMG profiles.

In vivo L3 validaGon (D10.4)

Brussels, 12th May 2016

Joint model: In vivo validaGon

Brussels, 12th May 2016

LEGEND

(Instrumented prosthesis data from Fregly et al., J Ort Res, 2012)

In vivo validaGon

RMSE = 0,36 (±0,05) RHO= 0,93 (±0,01)

Brussels, 12th May 2016

Ankle HerGan model

• Verifica6on: analyGcal model
• Sensi6vity: Primary source of uncertainty is reproducibility
• f joint geometric fixng (Prinold 2016)
• Valida6on - in vitro: extensive validaGon available in the

literature

• Valida6on - ex vivo: contact pressure esGmated compared

to those measured in cadaver experiments (L1 evidence)

• Confirma6on - clinics: on-going, provided by final data

analysis that correlated biomechanics predictors with clinical outcomes

Brussels, 12th May 2016