1 of 23

University of Michigan Quantitative Co-Clinical Imaging Research Resource

Brian D. Ross, Gary D. Luker, Moshe Talpaz & Thomas L. Chenevert

May 3^rd, 2022 CIRP Meeting Update

Objectives: To conduct a co-clinical trial to develop quantitative MRI biomarkers for myelofibrosis associated bone marrow phenotypes and disseminate results to the radiological and hematological/oncology communities.

Bone Marrow Fibrosis

Ineffective Hematopoiesis with Hypercellular Bone Marrow

Splenomegaly

2 of 23

The Village

Tom Chenevert

Brian Ross

Gary Luker

Moshe Talpaz

Kristen Pettit

Tim Johnson

Winston Lee

Amanda Fair

Tania Cunningham

Carrie McDonnell

Sanjee Palagama

Youngsoon Jang

Kevin Heist

Tanner Robinson

Dariya Malyarenko

NIH / NCI

U24CA237683

Christopher Bonham

4 of 23

https://doi.org/10.2147/IJGM.S51800

5 of 23

Jamieson et al. J Transl Med (2015) 13:294 DOI 10.1186/s12967-015-0644-4

Myelofibrosis Grading Criteria

6 of 23

Current State of MRI in MF

Quantify spleen volume for clinical trials
35% reduction in spleen volume used as end point for efficacy
Minimal/no applications in clinical practice
Initial bone marrow biopsy
Ruxolitinib FDA-approved Tx

7 of 23

Clinical Trial NCT01973881�Quantitative MRI for Myelofibrosis�PI. Gary Luker, MD, PhD.

Study Design:

Prospective
Observational - MRI:

Spleen Volume
ADC
PDFF
MTR

Longitudinal
Target recruitment 120 MF subjects
Recruitment to date 63 subjects, though lost timepoints due to Covid

Research Scans for Treatment Monitoring
Time of Research MRI	Time Relative to Treatment
Pre Treatment baseline	Within 4 weeks of starting treatment
After 1 month of therapy	+/- 2 weeks
After 3 months of therapy	+/- 4 weeks
After 6 months of therapy	+/- 4 weeks
Year 1 (optional)	+/- 3 months
Year 2 (optional)	+/- 3 months

8 of 23

Preliminary Clinical MR Data

MRI Observable Changes During Therapy

Patient with MF-2 fibrosis and 100% cellularity by bone marrow biopsy before treatment with ruxolitinib (Jak2-V617F driver mutation).

(Top) MIP reveals progressive reductions in bone marrow signal and spleen size on high b-value DWI.

(Mid) MR images and pseudocolor map of PDFF at weeks 0 and 24 of therapy. Yellow arrows show increased femoral head fat after therapy.

(Bottom) MTR values (calculated as (MToff – MTon)/MToff)). Red arrows show focal areas of high MTR at week 0 in iliac crests that decrease after 24 weeks of treatment. Note overall heterogeneity of MTR in iliac crests, revealing heterogeneity of bone marrow fibrosis that would be undetectable by biopsy.

9 of 23

Parametric Response Map (PRM) Analysis�Altered Spatial Distribution of Fat and Fibrosis Over 1 Year

Patient (MPLW515L mutation) had MF-3 with fibrosis and 95% cellularity by biopsy before starting therapy.

PRM PDFF displays increased pelvic and proximal femur fat with lesser changes in the posterior iliac crests, the site of bone marrow biopsy (lower image).

Scatterplot reveals 63.3% of all analyzed voxels (red) with increased fat exceeding confidence limits (angled lines).

Purple arrow points to a region in the left femoral head that did not exhibit increased fat, highlighting heterogeneity of treatment response.

PRM of MTR data show heterogeneous decreases in fibrosis (34.1% of all voxels with a decrease, blue). Depending on the selected site (right versus left iliac crest), oncologists would identify substantially different amounts of residual fibrosis.

10 of 23

University of Michigan U24 CA237683�Co-Clinical QIBs of Myelofibrosis: ADC, MTR, PDFF

WORKFLOW

2dseq MHD DICOM MHD DICOM MHD

Preclinical

Clinical

Phantoms

MHD QIB maps

Registration/

Segmentation/

Calibration

QIB stats

vs pathology

Acquisition

(SOP, Meta):

ADC(μm²/ms)

FatFr(%)

Hi-b MIP

MTR seg.

DATA FORMATS

FatFrac

ADC

MTR

path & endpoints

QA: accuracy,

repeatability

cellularity

ADC

histo

11 of 23

Bridge Preclinical & Clinical Phantoms *

Tube	MTR
1: 2mm	0.80 ± 0.02
2: 0.5mm	0.78 ± 0.06

Tube	ADC (μm²/ms)
3: 2mm	0.41 ± 0.02
4: 0.5mm	0.41 ± 0.01

Consistent quantities between large & small targets (≈0.5mm φ mouse tibia)

* Amouzandeh G. et al. ISMRM 2021 – CIRP Symposium

7T Preclinical 3T clinical

PDFF

	Tube	PDFF (%)
Solution 1	1: 2mm	28.07 ± 2.61
Solution 1	2: 0.5mm	33.10 ± 2.13
Solution 2	3: 2mm	7.97 ±1 .48
Solution 2	4: 0.5mm	6.61 ± 4.29

	Tube	PDFF (%)
Solution 1	4	27.21 ± 1.05
Solution 1	5	28.35 ± 0.76
Solution 2	6	5.03 ± 0.6

12 of 23

Bone Marrow Transplant Model of MF

Expected Changes in MRI Metrics

Myeloablative Irradiation

HSCs

Bone Marrow Harvest

Bone Marrow Transplant

MPL-W515L
JAK2-V617F
CALdel52
GFP (control)

(GFP co-expression)

Retrovirus Transduction

Reconstitution with Malignant HSCs

Enrich with CD117 beads

Spleen Volumetric MR Tibial MRI

13 of 23

Reproducibility: Inter-Mouse

Scans of 18 different healthy mice: VOI pixel histograms

Verticl

MTR

ADC

Fat Percent

Vertical lines denote 95% limits

14 of 23

Bland-Altman Test-Retest Intra-Mouse Repeatability of VOI Mean

Good precision for ADC & MTR: 95%LOA(ADC, MTR)<mean(ADC,MTR)
Moderate PDFF precision: LOA(FF) ~ mean(FF)
Negligible measurement bias

15 of 23

Anticipated Histological Correlations with MR Metrics

MRI Metric	Healthy	Hypercellular	Fibrosis (MF 1-3)
Fat Percent	Baseline	↓	↓
ADC	Baseline	↑	↑/↑/↑
MTR	Baseline	↑	↑/↑↑/↑↑↑

16 of 23

PDFF Results & Comparison to Histology :

Correlation to changes in cellular vs fat content over disease progression

Normal T1 T5

Distal

Proximal

Distal

180 μm

17 of 23

MRI Biomarkers Monitor Response to Ruxolitinib Therapy

18 of 23

https://umu24cirp.med.umich.edu/home

20 of 23

Data Repository

Subject-level metadata

21 of 23

Evaluation of Experimental Therapeutics

Nature Communications, In press, 2022.

22 of 23