Improving the Generalizability of Deep Learning for T2-Lesion Segmentation of Gliomas in the Post-Treatment Setting.

Although fully automated volumetric approaches for monitoring brain tumor response have many advantages, most available deep learning models are optimized for highly curated, multi-contrast MRI from newly diagnosed gliomas, which are not representative of post-treatment cases in the clinic. Improving segmentation for treated patients is critical to accurately tracking changes in response to therapy. We investigated mixing data from newly diagnosed (n = 208) and treated (n = 221) gliomas in training, applying transfer learning (TL) from pre- to post-treatment imaging domains, and incorporating spatial regularization for T2-lesion segmentation using only T2 FLAIR images as input to improve generalization post-treatment. These approaches were evaluated on 24 patients suspected of progression who had received prior treatment. Including 26% of treated patients in training improved performance by 13.9%, and including more treated and untreated patients resulted in minimal changes. Fine-tuning with treated glioma improved sensitivity compared to data mixing by 2.5% (p < 0.05), and spatial regularization further improved performance when used with TL by 95th HD, Dice, and sensitivity (6.8%, 0.8%, 2.2%; p < 0.05). While training with ≥60 treated patients yielded the majority of performance gain, TL and spatial regularization further improved T2-lesion segmentation to treated gliomas using a single MR contrast and minimal processing, demonstrating clinical utility in response assessment.

Improving the Automatic Classification of Brain MRI Acquisition Contrast with Machine Learning.

Automated quantification of data acquired as part of an MRI exam requires identification of the specific acquisition of relevance to a particular analysis. This motivates the development of methods capable of reliably classifying MRI acquisitions according to their nominal contrast type, e.g., T1 weighted, T1 post-contrast, T2 weighted, T2-weighted FLAIR, proton-density weighted. Prior studies have investigated using imaging-based methods and DICOM metadata-based methods with success on cohorts of patients acquired as part of a clinical trial. This study compares the performance of these methods on heterogeneous clinical datasets acquired with many different scanners from many institutions. RF and CNN models were trained on metadata and pixel data, respectively. A combined RF model incorporated CNN logits from the pixel-based model together with metadata. Four cohorts were used for model development and evaluation: MS research (n = 11,106 series), MS clinical (n = 3244 series), glioma research (n = 612 series, test/validation only), and ADNI PTSD (n = 477 series, training only). Together, these cohorts represent a broad range of acquisition contexts (scanners, sequences, institutions) and subject pathologies. Pixel-based CNN and combined models achieved accuracies between 97 and 98% on the clinical MS cohort. Validation/test accuracies with the glioma cohort were 99.7% (metadata only) and 98.4 (CNN). Accurate and generalizable classification of MRI acquisition contrast types was demonstrated. Such methods are important for enabling automated data selection in high-throughput and big-data image analysis applications.

Deep Learning to Predict Geographic Atrophy Area and Growth Rate from Multimodal Imaging.

OBJECTIVE: To develop deep learning models for annualized geographic atrophy (GA) growth rate prediction using fundus autofluorescence (FAF) images and spectral-domain OCT volumes from baseline visits, which can be used for prognostic covariate adjustment to increase power of clinical trials. DESIGN: This retrospective analysis estimated GA growth rate as the slope of a linear fit on all available measurements of lesion area over a 2-year period. Three multitask deep learning models-FAF-only, OCT-only, and multimodal (FAF and OCT)-were developed to predict concurrent GA area and annualized growth rate. PARTICIPANTS: Patients were from prospective and observational lampalizumab clinical trials. METHODS: The 3 models were trained on the development data set, tested on the holdout set, and further evaluated on the independent test sets. Baseline FAF images and OCT volumes from study eyes of patients with bilateral GA (NCT02247479; NCT02247531; and NCT02479386) were split into development (1279 patients/eyes) and holdout (443 patients/eyes) sets. Baseline FAF images from study eyes of NCT01229215 (106 patients/eyes) and NCT02399072 (169 patients/eyes) were used as independent test sets. MAIN OUTCOME MEASURES: Model performance was evaluated using squared Pearson correlation coefficient (r2) between observed and predicted lesion areas/growth rates. Confidence intervals were calculated by bootstrap resampling (B = 10 000). RESULTS: On the holdout data set, r2 (95% confidence interval) of the FAF-only, OCT-only, and multimodal models for GA lesion area prediction was 0.96 (0.95-0.97), 0.91 (0.87-0.95), and 0.94 (0.92-0.96), respectively, and for GA growth rate prediction was 0.48 (0.41-0.55), 0.36 (0.29-0.43), and 0.47 (0.40-0.54), respectively. On the 2 independent test sets, r2 of the FAF-only model for GA lesion area was 0.98 (0.97-0.99) and 0.95 (0.93-0.96), and for GA growth rate was 0.65 (0.52-0.75) and 0.47 (0.34-0.60). CONCLUSIONS: We show the feasibility of using baseline FAF images and OCT volumes to predict individual GA area and growth rates using a multitask deep learning approach. The deep learning-based growth rate predictions could be used for covariate adjustment to increase power of clinical trials. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found after the references.

Improving the noninvasive classification of glioma genetic subtype with deep learning and diffusion-weighted imaging.

BACKGROUND: Diagnostic classification of diffuse gliomas now requires an assessment of molecular features, often including IDH-mutation and 1p19q-codeletion status. Because genetic testing requires an invasive process, an alternative noninvasive approach is attractive, particularly if resection is not recommended. The goal of this study was to evaluate the effects of training strategy and incorporation of biologically relevant images on predicting genetic subtypes with deep learning. METHODS: Our dataset consisted of 384 patients with newly diagnosed gliomas who underwent preoperative MRI with standard anatomical and diffusion-weighted imaging, and 147 patients from an external cohort with anatomical imaging. Using tissue samples acquired during surgery, each glioma was classified into IDH-wildtype (IDHwt), IDH-mutant/1p19q-noncodeleted (IDHmut-intact), and IDH-mutant/1p19q-codeleted (IDHmut-codel) subgroups. After optimizing training parameters, top performing convolutional neural network (CNN) classifiers were trained, validated, and tested using combinations of anatomical and diffusion MRI with either a 3-class or tiered structure. Generalization to an external cohort was assessed using anatomical imaging models. RESULTS: The best model used a 3-class CNN containing diffusion-weighted imaging as an input, achieving 85.7% (95% CI: [77.1, 100]) overall test accuracy and correctly classifying 95.2%, 88.9%, 60.0% of the IDHwt, IDHmut-intact, and IDHmut-codel tumors. In general, 3-class models outperformed tiered approaches by 13.5%-17.5%, and models that included diffusion-weighted imaging were 5%-8.8% more accurate than those that used only anatomical imaging. CONCLUSION: Training a classifier to predict both IDH-mutation and 1p19q-codeletion status outperformed a tiered structure that first predicted IDH-mutation, then 1p19q-codeletion. Including apparent diffusion coefficient (ADC), a surrogate marker of cellularity, more accurately captured differences between subgroups.

Recurrent tumor and treatment-induced effects have different MR signatures in contrast enhancing and non-enhancing lesions of high-grade gliomas.

BACKGROUND: Differentiating treatment-induced injury from recurrent high-grade glioma is an ongoing challenge in neuro-oncology, in part due to lesion heterogeneity. This study aimed to determine whether different MR features were relevant for distinguishing recurrent tumor from the effects of treatment in contrast-enhancing lesions (CEL) and non-enhancing lesions (NEL). METHODS: This prospective study analyzed 291 tissue samples (222 recurrent tumor, 69 treatment-effect) with known coordinates on imaging from 139 patients who underwent preoperative 3T MRI and surgery for a suspected recurrence. 8 MR parameter values were tested from perfusion-weighted, diffusion-weighted, and MR spectroscopic imaging at each tissue sample location for association with histopathological outcome using generalized estimating equation models for CEL and NEL tissue samples. Individual cutoff values were evaluated using receiver operating characteristic curve analysis with 5-fold cross-validation. RESULTS: In tissue samples obtained from CEL, elevated relative cerebral blood volume (rCBV) was associated with the presence of recurrent tumor pathology (P < 0.03), while increases in normalized choline (nCho) and choline-to-NAA index (CNI) were associated with the presence of recurrent tumor pathology in NEL tissue samples (P < 0.008). A mean CNI cutoff value of 2.7 had the highest performance, resulting in mean sensitivity and specificity of 0.61 and 0.81 for distinguishing treatment-effect from recurrent tumor within the NEL. CONCLUSION: Although our results support prior work that underscores the utility of rCBV in distinguishing the effects of treatment from recurrent tumor within the contrast enhancing lesion, we found that metabolic parameters may be better at differentiating recurrent tumor from treatment-related changes in the NEL of high-grade gliomas.

Blood vessel anastomosis is spatially regulated by Flt1 during angiogenesis.

Blood vessel formation is essential for vertebrate development and is primarily achieved by angiogenesis - endothelial cell sprouting from pre-existing vessels. Vessel networks expand when sprouts form new connections, a process whose regulation is poorly understood. Here, we show that vessel anastomosis is spatially regulated by Flt1 (VEGFR1), a VEGFA receptor that acts as a decoy receptor. In vivo, expanding vessel networks favor interactions with Flt1 mutant mouse endothelial cells. Live imaging in human endothelial cells in vitro revealed that stable connections are preceded by transient contacts from extending sprouts, suggesting sampling of potential target sites, and lowered Flt1 levels reduced transient contacts and increased VEGFA signaling. Endothelial cells at target sites with reduced Flt1 and/or elevated protrusive activity were more likely to form stable connections with incoming sprouts. Target cells with reduced membrane-localized Flt1 (mFlt1), but not soluble Flt1, recapitulated the bias towards stable connections, suggesting that relative mFlt1 expression spatially influences the selection of stable connections. Thus, sprout anastomosis parameters are regulated by VEGFA signaling, and stable connections are spatially regulated by endothelial cell-intrinsic modulation of mFlt1, suggesting new ways to manipulate vessel network formation.

Flt-1 (VEGFR-1) coordinates discrete stages of blood vessel formation.

AIMS: In developing blood vessel networks, the overall level of vessel branching often correlates with angiogenic sprout initiations, but in some pathological situations, increased sprout initiations paradoxically lead to reduced vessel branching and impaired vascular function. We examine the hypothesis that defects in the discrete stages of angiogenesis can uniquely contribute to vessel branching outcomes. METHODS AND RESULTS: Time-lapse movies of mammalian blood vessel development were used to define and quantify the dynamics of angiogenic sprouting. We characterized the formation of new functional conduits by classifying discrete sequential stages-sprout initiation, extension, connection, and stability-that are differentially affected by manipulation of vascular endothelial growth factor-A (VEGF-A) signalling via genetic loss of the receptor flt-1 (vegfr1). In mouse embryonic stem cell-derived vessels genetically lacking flt-1, overall branching is significantly decreased while sprout initiations are significantly increased. Flt-1(-/-) mutant sprouts are less likely to retract, and they form increased numbers of connections with other vessels. However, loss of flt-1 also leads to vessel collapse, which reduces the number of new stable conduits. Computational simulations predict that loss of flt-1 results in ectopic Flk-1 signalling in connecting sprouts post-fusion, causing protrusion of cell processes into avascular gaps and collapse of branches. Thus, defects in stabilization of new vessel connections offset increased sprout initiations and connectivity in flt-1(-/-) vascular networks, with an overall outcome of reduced numbers of new conduits. CONCLUSIONS: These results show that VEGF-A signalling has stage-specific effects on vascular morphogenesis, and that understanding these effects on dynamic stages of angiogenesis and how they integrate to expand a vessel network may suggest new therapeutic strategies.