The Clotbust Review

DWI-Positive Lesions in Patients With Acute Intracerebral Hemorrhage — Associations With Cerebral Small Vessel Disease and Clinical Outcome

· DOI: 10.1212/wnl.0000000000218449 · PMC13528906 · stroke deep-dive intracerebral hemorrhage diffusion-weighted imaging cerebral small vessel disease

Stylized illustration of DWI-positive lesions in acute intracerebral hemorrhage
AI-generated editorial illustration. Generated for editorial use.

Clinical Question (PICO)

Population: Adults with imaging-confirmed acute spontaneous intracerebral hemorrhage (ICH) who had diagnostic-quality MRI during the acute admission. Exposure/comparator: remote DWI-positive lesions versus no DWI-positive lesions; no treatment was assigned. Outcomes: prevalence and anatomical distribution of DWI-positive lesions, associations with cerebral small-vessel-disease (cSVD) markers, poor six-month functional outcome (modified Rankin Scale [mRS] 3–6), and recurrent stroke. A planned etiologic comparison examined probable cerebral amyloid angiopathy (CAA) versus non-CAA ICH; there was no randomized treatment subgroup.

Bottom Line

Remote DWI-positive lesions were present in 74 of 342 patients (21.6%) and were independently associated with poor six-month function (adjusted OR 2.16, 95% CI 1.67–3.48; p < 0.001), as well as severe white-matter hyperintensities, larger hematoma volume, and greater cerebral microbleed burden. The lesions are a promising marker of active or severe cSVD and prognosis after ICH, but this observational study does not establish that they cause disability or identify a treatment response.

Design

  • Trial type: Prospective observational cohort using consecutive patients from the Stroke Investigation Group in North and Central London (SIGNAL) registry; no randomization or assigned intervention.
  • N: 342 included; 332 had six-month outcome data
    • DWI-positive: 74
    • DWI-negative: 268
  • Randomization: None. Groups were defined by MRI exposure status; CAA versus non-CAA distribution was a planned etiologic comparison.
  • Setting: University College London Hospitals NHS Foundation Trust, serving the approximately 1.5-million-person North Central London region.
  • Enrollment: January 1, 2017, through January 13, 2020.
  • Mean follow-up: Six months for functional outcome and recurrent stroke; MRI was obtained a median of 4 days after admission (IQR 2–9).
  • Analysis: Univariable comparisons followed by multivariable logistic regression. Candidate variables were selected with least absolute shrinkage and selection operator methods; age, sex, and stroke severity were retained as clinically relevant covariates. Results were reported as ORs with 95% CIs.
  • Primary outcome: Unfavorable functional outcome at six months, defined as mRS 3–6 (dependence or death).

Population

Inclusion Criteria

  • Age older than 18 years.
  • Imaging-confirmed acute ICH recorded in the SIGNAL registry.
  • Diagnostic-quality brain MRI, including DWI and apparent diffusion coefficient maps, during the acute hospital admission and within 90 days of emergency admission.

Exclusion Criteria

  • No diagnostic-quality MRI or no DWI sequence.
  • Isolated intraventricular hemorrhage or subarachnoid hemorrhage.
  • Secondary ICH caused by a macrovascular lesion, tumor, or other structural lesion.

Baseline Characteristics (overall or representative arm)

  • Median age 67 years (IQR 55–78); 142 participants (41.5%) were women.
  • Median admission NIH Stroke Scale score was 6 (IQR 4–12).
  • ICH location was deep in 162 patients (47.4%), lobar in 138 (40.3%), infratentorial in 32 (9.4%), and mixed in 10 (2.9%).
  • Median hematoma volume was 9 mL (IQR 3–23), and median total cSVD burden score was 2 (IQR 1–3).
  • Probable CAA was present in 64 patients (18.7%); arteriolosclerosis in 131 (38.3%); mixed SVD in 126 (36.8%); and no MRI-visible SVD in 21 (6.1%).

Interventions

  • DWI-positive exposure: At least one DWI-hyperintense lesion with corresponding restricted diffusion, located more than 1 cm from the hematoma and classified as lobar, deep, infratentorial, or mixed.
  • DWI-negative comparator: No qualifying remote DWI-positive lesion on the acute MRI protocol.
  • Clinical treatment: No study treatment was assigned. Acute blood-pressure management, surgery, hemostatic therapy, and secondary prevention were not randomized by this study.

Outcomes

Primary Outcome (mRS 3–6 at six months):

  • Follow-up was available for 332 patients; 198 had mRS 3–6.
  • DWI-positive lesions were present in 26.26% of patients with poor outcome versus 14.92% of those with favorable outcome.
  • Median six-month mRS was 3 (IQR 2–5) in the DWI-positive group versus 2 (IQR 1–3) in the DWI-negative group (p < 0.001).
  • After adjustment, DWI-positive lesions remained associated with unfavorable outcome (OR 2.16, 95% CI 1.67–3.48; p < 0.001). This was an observational prognostic association, not a treatment effect.

Secondary Outcomes:

  • DWI-positive lesions occurred in 74 of 342 patients (21.6%). Among positive scans, 38 patients (51.4%) had one lesion, 24 (32.4%) had two or three, 7 (9.5%) had four or five, and 5 (6.8%) had at least six.
  • Lesions were most often in white matter (39/74, 52.7%) and were strictly lobar in 33/74 (44.6%), strictly deep in 15/74 (20.3%), strictly infratentorial in 5/74 (6.7%), and mixed in 21/74 (28.4%).
  • Independent correlates of DWI positivity were prior stroke or TIA (OR 1.84, 95% CI 1.47–3.37; p < 0.001), ICH volume at least 30 mL (OR 2.29, 95% CI 1.76–3.42; p < 0.001), Fazekas grade at least 2 white-matter hyperintensities (OR 3.37, 95% CI 2.72–6.89; p < 0.001), and at least five cerebral microbleeds (OR 1.61, 95% CI 1.21–2.93; p = 0.007).
  • Overall DWI-positive prevalence did not differ between CAA and non-CAA groups. In the small DWI-positive CAA subgroup, lesions were more often strictly lobar than in non-CAA disease (78.6% vs 36.7%), while strictly deep lesions occurred only in the non-CAA group (0% vs 25.0%). This etiologic comparison was hypothesis generating.
  • Recurrent stroke at six months was more frequent with DWI-positive lesions (20.3% vs 6.3%; p = 0.037), including ischemic stroke (6.8% vs 3.4%; p = 0.016) and recurrent ICH (13.5% vs 3.0%; p = 0.001). These were unadjusted event comparisons with small counts.

Adverse Events / Safety:

  • No intervention was administered, so treatment-related bleeding or other adverse events were not assessed.
  • Six-month mortality was 13.5% in the DWI-positive group versus 4.8% in the DWI-negative group; the difference did not meet the study’s statistical-significance threshold (p = 0.062).

Figures

MRI showing remote DWI-positive lesions after lobar intracerebral hemorrhage
Figure 1. MRI findings in a representative patient with lobar ICH. DWI sequences show the parietal hematoma and multiple remote foci of restricted diffusion in lobar, deep, and white-matter locations; FLAIR shows early confluent periventricular white-matter hyperintensities.

Source: PMC PMC13528906 — WNL-2025-207367f1.webp. Click image to expand.

Six-month modified Rankin Scale distribution by DWI lesion status
Figure 2. Six-month modified Rankin Scale distribution in the DWI-positive and DWI-negative groups. Follow-up data were unavailable for 10 patients: 2 with DWI-positive lesions and 8 without.

Source: PMC PMC13528906 — WNL-2025-207367f2.webp. Click image to expand.

Criticisms

  • MRI availability created substantial selection pressure: 205 of 547 screened patients were excluded, including 143 with CT-only imaging because of contraindications, greater stroke severity, dementia, or instability. Excluded patients were older and more medically fragile, so the cohort likely underrepresents the most severe ICH population.
  • MRI was permitted within a 90-day window, although the median was 4 days. DWI lesions can appear and evolve over time, and one scan cannot determine whether a lesion preceded the hemorrhage, followed acute blood-pressure treatment, or arose from the same active arteriopathy.
  • The study is observational. DWI positivity may be a marker of hematoma severity and cSVD burden rather than a causal contributor to disability; residual confounding remains despite regression adjustment and LASSO variable selection.
  • Detailed hyperacute blood-pressure-lowering, surgical, and hemostatic-treatment data were unavailable, limiting assessment of potentially modifiable mechanisms.
  • The recurrent-stroke comparisons were unadjusted and based on small event counts. The study did not report a time-to-event model or externally validate DWI positivity as an incremental prediction tool beyond established ICH severity measures.
  • Only 14 patients with probable CAA had DWI-positive lesions. The absence of deep lesions in that small subgroup cannot yet be used to exclude CAA or revise diagnostic criteria.
  • This was a single regional registry with no external validation, and six-month outcomes were missing for 10 patients.

Funding

The National Institute for Health and Care Research University College London Hospitals Biomedical Research Centre funded the research and supported the design of the SIGNAL study and data collection. The authors reported no relevant disclosures. There was no clinical-trial registration; the SIGNAL registry operated under UCL Hospitals governance approval 5-201920-SE and London South-East Research Ethics Committee approval 24/LO/0368, with individual consent waived for routinely collected data.

The paper

  • Authors. Locatelli et al.
  • Title. DWI-Positive Lesions in Patients With Acute Intracerebral Hemorrhage: Associations With Cerebral Small Vessel Disease and Clinical Outcome.
  • Journal. Neurology.
  • Year. 2026.
  • DOI. 10.1212/wnl.0000000000218449
  • PMCID. PMC13528906
Deep Dive — click to expand

What this is

This is a prospective registry analysis asking whether small, remote restricted-diffusion lesions seen after spontaneous ICH are incidental imaging noise or a clinically meaningful signature of active cerebral small-vessel disease. The answer is directionally strong: one in five MRI-selected patients had a DWI-positive lesion, and positivity tracked with severe white-matter disease, larger hematomas, more microbleeds, worse six-month function, and more recurrent stroke. The study supports reading and reporting these lesions as prognostic information; it does not prove that preventing the lesions, changing acute blood-pressure targets, or altering antithrombotic decisions will improve outcome.

1. Shadow Audit

The headline adjusted OR of 2.16 can obscure what the scan is actually measuring. DWI positivity clustered with prior stroke or TIA, hematoma volume at least 30 mL, severe white-matter hyperintensities, and at least five microbleeds, so it may be a compact image of pre-existing disease burden plus index-ICH severity rather than an independent pathophysiologic driver. The cohort also lost 205 of 547 screened patients, chiefly because the older, sicker, or unstable patients could not undergo MRI; that 37.5% exclusion fraction makes “consecutive registry” less representative than it first sounds. The paper additionally lacked detailed hyperacute blood-pressure-treatment data, leaving unresolved whether some lesions reflected underlying vasculopathy, acute hemodynamic stress, treatment exposure, or all three.

2. Inversion Engine

For the opposite conclusion—“remote DWI lesions add no useful prognostic information beyond routine ICH variables”—to hold, external validation would have to shrink the adjusted association from OR 2.16 to approximately 1.0, with the lower 95% confidence limit moving from 1.67 to below 1 after complete adjustment for baseline disability, NIHSS, hematoma volume, cSVD burden, treatment, and MRI timing. The observed unadjusted outcome separation is clinically visible but not enormous: the reported counts imply poor outcome in roughly 72% of DWI-positive versus 56% of DWI-negative patients with follow-up, an absolute difference near 16 percentage points. If a better-specified model reduced that difference below about 5 points and showed no gain in discrimination or calibration, DWI positivity would remain mechanistically interesting but cease to be a useful bedside risk marker.

3. Second-Order Catalyst

The first institutions to change should be tertiary ICH services that already obtain MRI and run dedicated hemorrhage or CAA clinics. The immediate protocol change is not a new drug: neuroradiology reports should explicitly identify remote restricted-diffusion lesions, state whether they are lobar, deep, infratentorial, or mixed, and pair that finding with the visible cSVD burden. Stroke teams can then route DWI-positive patients toward deliberate six-month follow-up, recurrent-stroke surveillance, and a complete secondary-prevention review. This could be implemented in a reporting template within weeks; any change to acute blood-pressure lowering or antithrombotic timing should wait for interventional evidence.

4. Asymmetric Leverage

The asymmetry is that DWI is already part of many acute ICH MRI protocols, so extracting this signal costs little once the scan exists. A one-line report addition may identify the roughly 22% of MRI-selected patients in whom six-month recurrent stroke was 20.3% rather than 6.3% and poor outcome was more common. The payoff could therefore be meaningful across a large ICH denominator even if the lesion itself is not causal. The constraint is equally important: many of the sickest patients never receive MRI, so a DWI-based pathway cannot become the universal ICH prognostic system and should complement—not replace—NIHSS, hematoma volume, baseline function, and established cSVD markers.

5. Paradigm Destroyer

This paper weakens the reflex to dismiss tiny remote DWI lesions after ICH as incidental embolic debris or radiologic trivia. It also rejects the opposite reflex of treating them as an actionable treatment target before causality is established. Tomorrow morning’s protocol could read: “When MRI after spontaneous ICH shows restricted diffusion more than 1 cm from the hematoma, document lesion number and distribution, summarize cSVD burden, and flag the patient for structured six-month outcome and recurrence follow-up. Do not alter blood-pressure or antithrombotic strategy solely because the lesion is present.”

MVP — Minimum Viable Proof

The minimum practice-changing proof would be a blinded external prospective cohort in which DWI positivity adds reproducible prognostic value beyond age, premorbid mRS, NIHSS, hematoma location and volume, and cSVD burden—ideally confirming at least a 10-percentage-point absolute increase in six-month dependence/death or recurrent stroke and demonstrating improved calibration or net reclassification. A subsequent randomized or protocolized study would still be needed before using the marker to select a specific treatment.

Best Combination

Combine DWI positivity with conventional ICH severity measures and the rest of the MRI phenotype rather than using it alone. Hematoma volume and NIHSS describe the acute insult; white-matter hyperintensities, microbleeds, lacunes, and siderosis describe the chronic arteriopathy; remote DWI lesions may identify current ischemic activity within that substrate. The coherent bedside model is therefore a layered risk assessment: acute injury plus chronic cSVD burden plus active remote tissue injury, followed by etiologic classification into CAA, arteriolosclerosis, or mixed disease and longitudinal validation against functional and recurrent-stroke outcomes.

Overvalue Warning

  • Do not read OR 2.16 as a 2.16-fold absolute risk or as proof that the DWI lesion caused disability. The outcome was common, the cohort was MRI-selected, and residual confounding by ICH and cSVD severity is likely.
  • Do not use a deep DWI lesion to rule out CAA, or use DWI positivity alone to change acute blood-pressure or antithrombotic management. The CAA/DWI-positive subgroup contained only 14 patients, and no treatment interaction was tested.

Educational use notice. The Clotbust Review is dedicated to medical education. Figures from published articles are embedded for teaching purposes, with attribution to the original publication. We do not claim copyright over any embedded figure; all rights remain with the original publisher and authors under the article's published license. If you are a rights holder and would like a figure removed, please contact the editorial team. See our editorial notice for the full policy.