RESUMO
Improvements to patient care through the development of automated image analysis in pathology are restricted by the small image patch size that can be processed by convolutional neural networks (CNNs), when compared to the whole-slide image (WSI). Tile-by-tile processing across the entire WSI is slow and inefficient. While this may improve with future computing power, the technique remains vulnerable to noise from uninformative image areas. We propose a novel attention-inspired algorithm that selects image patches from informative parts of the WSI, first using a sparse randomised grid pattern, then iteratively re-sampling at higher density in regions where a CNN classifies patches as tumour. Subsequent uniform sampling across the enclosing region of interest (ROI) is used to mitigate sampling bias. Benchmarking tests informed the adoption of VGG19 as the main CNN architecture, with 79% classification accuracy. A further CNN was trained to separate false-positive normal epithelium from tumour epithelium, in a novel adaptation of a two-stage model used in brain imaging. These subsystems were combined in a processing pipeline to generate spatial distributions of classified patches from unseen WSIs. The ROI was predicted with a mean F1 (Dice) score of 86.6% over 100 evaluation WSIs. Several algorithms for evaluating tumour-stroma ratio (TSR) within the ROI were compared, giving a lowest root mean square (RMS) error of 11.3% relative to pathologists' annotations, against 13.5% for an equivalent tile-by-tile pipeline. Our pipeline processed WSIs between 3.3x and 6.3x faster than tile-by-tile processing. We propose our attention-based sampling pipeline as a useful tool for pathology researchers, with the further potential for incorporating additional diagnostic calculations.
RESUMO
OBJECTIVE The aim of this study was to investigate long-term seizure outcome, rate of reoperation, and postoperative neuropsychological performance following selective amygdalohippocampectomy (SelAH) or anterior temporal lobectomy (ATL) in pediatric patients with medically refractory temporal lobe epilepsy (TLE). METHODS The authors performed a retrospective review of cases of medically refractory pediatric TLE treated initially with either SelAH or ATL. Standardized pre- and postoperative evaluation included seizure charting, surface and long-term video-electroencephalography, 1.5-T MRI, and neuropsychological testing. RESULTS A total of 79 patients treated initially with SelAH (n = 18) or ATL (n = 61) were included in this study, with a mean follow-up of 5.3 ± 4 years (range 1-16 years). The patients' average age at initial surgery was 10.6 ± 5 years, with an average surgical delay of 5.7 ± 4 years between seizure onset and surgery. Seizure freedom (Engel I) following the initial operation was significantly more likely following ATL (47/61, 77%) than SelAH (8/18, 44%; p = 0.017, Fisher's exact test). There was no statistically significant difference in the proportion of patients with postoperative neuropsychological deficits following SelAH (8/18, 44%) or ATL (21/61, 34%). However, reoperation was significantly more likely following SelAH (8/18, 44%) than after ATL (7/61, 11%; p = 0.004) and was more likely to result in Engel I outcome for ATL after failed SelAH (7/8, 88%) than for posterior extension after failed ATL (1/7, 14%; p = 0.01). Reoperation was well tolerated without significant neuropsychological deterioration. Ultimately, including 15 reoperations, 58 of 79 (73%) patients were free from disabling seizures at the most recent follow-up. CONCLUSIONS SelAH among pediatric patients with medically refractory unilateral TLE yields significantly worse rates of seizure control compared with ATL. Reoperation is significantly more likely following SelAH, is not associated with incremental neuropsychological deterioration, and frequently results in freedom from disabling seizures. These results are significant in that they argue against using SelAH for pediatric TLE surgery.