Impact of Inversion Time for FLAIR Acquisition on the T2-FLAIR Mismatch Detectability for IDH-Mutant, Non-CODEL Astrocytomas

Manabu Kinoshita, Hideyuki Arita, Masamichi Takahashi, Takehiro Uda, Junya Fukai, Kenichi Ishibashi, Noriyuki Kijima, Ryuichi Hirayama, Mio Sakai, Atsuko Arisawa, Hiroto Takahashi, Katsuyuki Nakanishi, Naoki Kagawa, Kouichi Ichimura, Yonehiro Kanemura, Yoshitaka Narita, Haruhiko KishimaFrontiers in Oncology 2021 · DOI

TL;DR

This study demonstrates that optimizing the inversion time (TI) in FLAIR MRI sequences significantly improves the detection of the "T2-FLAIR mismatch sign," a crucial indicator for identifying IDH-mutant, non-CODEL astrocytomas. By using shorter TIs (less than 2,400 ms on 3T scanners), diagnostic accuracy, particularly sensitivity and negative predictive value, is substantially enhanced, making this a valuable tool for clinical practice.

What Problem Does This Paper Address?

Accurately classifying gliomas is critical for guiding treatment and predicting patient outcomes. Molecular markers like IDH mutation and 1p/19q co-deletion are key, but non-invasively identifying them pre-treatment is challenging. The "T2-FLAIR mismatch sign" has emerged as a promising imaging biomarker for IDH-mutant, non-CODEL astrocytomas, characterized by hyperintensity on T2-weighted images (T2WI) and relative hypointensity on FLAIR (except for a peripheral rim). However, the sensitivity of this sign has been a limitation, meaning some IDH-mutant, non-CODEL astrocytomas don't exhibit it. Previous research suggested that FLAIR acquisition parameters, specifically inversion time (TI), might influence the visibility of this mismatch. This paper aims to systematically investigate how TI affects the diagnostic accuracy of the T2-FLAIR mismatch sign and to determine an optimal TI for its reliable detection.

Methods

The researchers employed a multi-cohort approach to test their hypothesis. They analyzed MRI data from three distinct groups:

  1. Cohort 1: A local dataset of 94 MRI studies from 76 histologically and molecularly confirmed IDH-mutant, non-CODEL lower-grade gliomas (LrGGs) with varying FLAIR acquisition parameters.
  2. Cohort 2: A local dataset of 33 MRI studies from 31 LrGGs specifically acquired with FLAIR using a TI less than 2,400 ms on 3T scanners (or equivalent on 1.5T).
  3. Cohort 3: A larger dataset of 112 MRI studies from the TCIA/TCGA (The Cancer Imaging Archive/The Cancer Genome Atlas) for LrGGs, used for external validation.

The core of their methodology involved evaluating the presence or absence of the T2-FLAIR mismatch sign by two experienced neuroradiologists. They then statistically compared the diagnostic accuracies (sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV)) based on the TI used for FLAIR acquisition. They also used Receiver Operating Characteristic (ROC) curves to assess overall diagnostic performance. A novel aspect was their method for converting TI values between 1.5T and 3T scanners to ensure consistent comparisons, using MP2RAGE sequences and a derived conversion coefficient.

Datasets

  • Local Datasets:
    • Cohort 1: 94 MRI studies from 76 IDHmt, non-CODEL LrGGs.
  • Cohort 2: 33 MRI studies from 31 LrGGs with short TI-FLAIR.
  • Publicly Available Dataset:
    • Cohort 3: 112 MRI studies from the TCIA/TCGA dataset for LrGGs.

The study did not introduce a new benchmark dataset but rather utilized existing, well-established public repositories (TCIA/TCGA) for validation, which is a standard practice for robust scientific findings.

Key Insights and Ideas

  • TI is not just a parameter, it's a diagnostic lever: The study highlights that FLAIR acquisition parameters are not merely technical details but can profoundly impact the diagnostic utility of imaging biomarkers.
  • Shorter TI unlocks the mismatch: The core insight is that reducing the inversion time in FLAIR sequences (specifically below 2,400 ms on 3T scanners) significantly enhances the detectability of the T2-FLAIR mismatch sign. This suggests that the conventional FLAIR sequence, optimized for other conditions, might not be ideal for LrGG characterization.
  • Specificity remains high: While sensitivity is improved, the study also shows that the T2-FLAIR mismatch sign, even with shorter TIs, remains highly specific for IDH-mutant, non-CODEL astrocytomas, meaning a positive finding is a strong indicator of this specific tumor type.
  • Clinical translation is feasible: The T2-FLAIR mismatch sign, when acquired with optimized TI, offers a readily applicable imaging biomarker that clinicians can integrate into practice without requiring specialized or complex analysis techniques.

Key Results

  • Improved Detectability: In Cohort 1, the frequency of the T2-FLAIR mismatch sign was significantly higher (0.76 vs. 0.34) when FLAIR was acquired with TI shorter than 2,400 ms (p = 0.0009).
  • Specificity Confirmed: In Cohort 2, the T2-FLAIR mismatch sign was positive only for IDHmt, non-CODEL astrocytomas when using shorter TI (p = 0.0001), demonstrating its high specificity.
  • Validation in TCIA/TCGA: The TCIA/TCGA dataset (Cohort 3) confirmed these findings. Sensitivity improved from 31% to 67%, specificity from 90% to 94%, PPV from 79% to 92%, and NPV from 51% to 74% when using shorter TI.
  • ROC Curve Improvement: The area under the ROC curve for identifying IDHmt, non-CODEL astrocytomas improved from 0.63 to 0.87 with shorter TI.
  • Qualitative Observation: Visual inspection of cases with multiple FLAIR acquisitions (Figure 3) clearly showed that the T2-FLAIR mismatch sign was more apparent with shorter TI.

Limitations and Future Work

  • Sensitivity Still Moderate: Despite improvements, the sensitivity of the T2-FLAIR mismatch sign, even with shorter TI, is around 70%. This means an absence of the sign cannot definitively rule out IDHmt, non-CODEL astrocytomas.
  • Inter-rater Reliability: The qualitative nature of the sign led to variable inter-rater reliability (Cohen's kappa ranging from 0.32 to 0.60), suggesting that careful reading by experienced clinicians is still crucial.
  • External Validation Needed: While the TCIA/TCGA dataset provides external validation, the authors suggest that further validation in independent external cohorts would be beneficial.
  • Future Work: The authors propose that acquiring FLAIR with TI shorter than 2,400 ms on 3T scanners should become standard practice for LrGG imaging. Further research could explore the precise mechanisms behind the TI's influence and potentially refine TI values for even greater accuracy.

Summary updated

Abstract

The current research tested the hypothesis that inversion time (TI) shorter than 2,400 ms under 3T for FLAIR can improve the diagnostic accuracy of the T2-FLAIR mismatch sign for identifying <i>IDH</i>mt, non-CODEL astrocytomas. We prepared three different cohorts; 94 MRI from 76 <i>IDH</i>mt, non-CODEL Lower-grade gliomas (LrGGs), 33 MRI from 31 LrGG under the restriction of FLAIR being acquired with TI < 2,400 ms for 3T or 2,016 ms for 1.5T, and 112 MRI from 112 patients from the TCIA/TCGA dataset for LrGG. The presence or absence of the "T2-FLAIR mismatch sign" was evaluated, and we compared diagnostic accuracies according to TI used for FLAIR acquisition. The T2-FLAIR mismatch sign was more frequently positive when TI was shorter than 2,400 ms under 3T for FLAIR acquisition (<i>p</i> = 0.0009, Fisher's exact test). The T2-FLAIR mismatch sign was positive only for <i>IDH</i>mt, non-CODEL astrocytomas even if we confined the cohort with FLAIR acquired with shorter TI (<i>p</i> = 0.0001, Fisher's exact test). TCIA/TCGA dataset validated that the sensitivity, specificity, PPV, and NPV of the T2-FLAIR mismatch sign to identify <i>IDH</i>mt, non-CODEL astrocytomas improved from 31, 90, 79, and 51% to 67, 94, 92, and 74%, respectively and the area under the curve of ROC improved from 0.63 to 0.87 when FLAIR was acquired with shorter TI. We revealed that TI for FLAIR impacts the T2-FLAIR mismatch sign's diagnostic accuracy and that FLAIR scanned with TI < 2,400 ms in 3T is necessary for LrGG imaging.
BibTeX
@inproceedings{kinoshita2021impact,
  author = {Manabu Kinoshita and Hideyuki Arita and Masamichi Takahashi and Takehiro Uda and Junya Fukai and Kenichi Ishibashi and Noriyuki Kijima and Ryuichi Hirayama and Mio Sakai and Atsuko Arisawa and Hiroto Takahashi and Katsuyuki Nakanishi and Naoki Kagawa and Kouichi Ichimura and Yonehiro Kanemura and Yoshitaka Narita and Haruhiko Kishima},
  title = {{Impact of Inversion Time for FLAIR Acquisition on the T2-FLAIR Mismatch Detectability for IDH-Mutant, Non-CODEL Astrocytomas}},
  booktitle = {Frontiers in Oncology},
  year = {2021},
  doi = {10.3389/fonc.2020.596448},
}
Figure