Improving Multiple Sclerosis Plaque Detection Using a Semiautomated Assistive Approach

J. van Heerden; D. Rawlinson; A.M. Zhang; R. Chakravorty; M.A. Tacey; P.M. Desmond; F. Gaillard

doi:10.3174/ajnr.A4375

Article Figures & Data

Figures

Tables

Download figure
Open in new tab
Download powerpoint
Fig 1.
Annotated capture of the software reporting screen. A, Axial FLAIR with superimposed change map shows the new occipital white matter lesion in orange. Coregistered and resectioned FLAIR sequences comparing axial of new study (B) with axial of old study (C); and sagittal of new study (E) with sagittal old study (F)—thus confirming that the lesion is real and consistent with a new demyelinating plaque. D, Each lesion is marked with 3D coordinates.
Download figure
Open in new tab
Download powerpoint
Fig 2.
Preprocessing for change-detection on receipt of a new study. A pair of old and new studies are required, each containing a volumetric series used for change detection. In our case, this series uses the FLAIR protocol. Due to significant deformation in soft tissues outside the cranium, it is preferable to register the studies by using only the brain tissue. To this end, a brain-surface extraction tool (BrainSuite from the University of Southern California)¹³ is fitted (1) and then used to mask the brain in the new study (2). Next, the equivalent series in the old study is retrieved and coregistered to the new study (3) by using the Mutual Information algorithm. The recovered transformation is stored in the PACS data base. Note that it is only necessary to mask the new study during registration and that rigid registration yielded sufficient accuracy after exclusion of the masked areas. DOF indicates degrees of freedom.
Download figure
Open in new tab
Download powerpoint
Fig 3.
A, Comparative graphic representation of the number of study pairs with new lesions detected by both readers when using the software compared to the issued radiology report. B, Comparing the number of study pairs improved with demyelinating lesions detected by both readers when using the newly developed assistive software to the issued radiology report.

Tables

Figures

View popup
Table 1:
Demonstrating the number of study pairs showing a change in lesion load as identified using conventional side-by-side comparison and the software
Change in Lesion Load Issued Radiology Report Reader 1 Reader 2
Study pairs with new lesions (No.) 20 60 (P < .001) 62 (P < .001)
Study pairs with improved lesions (No.) 5 28 (P < .001) 39 (P < .001)

View popup

Table 2:

Interreader agreement demonstrated with binary groupings of new and improved lesions when using the software

Change in Lesion Load, Binary Grouping	κ	95% CI
New lesions (0, 1+)	0.87	0.79–0.95
New lesions (0–1, 2+)	0.81	0.71–0.91
New lesions (0–2, 3+)	0.96	0.90–1.00
Improved lesions (0, 1+)	0.72	0.59–0.85
Improved lesions (0–1, 2+)	0.79	0.64–0.94
Improved lesions (0–2, 3+)	0.70	0.49–0.91

View popup

Table 3:

Intrareader agreement demonstrated with binary groupings of new and improved lesions using both conventional side-by-side comparison and the software^{^a}

	New Lesions (κ) (95% CI)	Improved Lesions (κ) (95% CI)
One or more lesions
VTS 1st vs VTS 2nd read	1.000	0.937 (0.815–1.000)
CSSC 1st vs CSSC 2nd read	0.941 (0.826–1.000)	0.462 (0.039–0.886)
Two or more lesions
VTS 1st vs VTS 2nd read	1.000	0.731 (0.448–1.000)
CSSC 1st vs CSSC 2nd read	0.846 (0.640–1.000)	0.482 (−0.118–1.000)
Three or more lesions
VTS 1st vs VTS 2nd read	1.000	0.774 (0.472–1.000)
CSSC 1st vs CSSC 2nd read	0.724 (0.361–1.000)	0.482 (−0.118–1.000)

↵a Correlations demonstrated substantial intrareader agreement. The software generally outperformed conventional side-by-side comparison without, however, reaching statistical significance.