TRUSformer: improving prostate cancer detection from micro-ultrasound using attention and self-supervision.

PURPOSE: A large body of previous machine learning methods for ultrasound-based prostate cancer detection classify small regions of interest (ROIs) of ultrasound signals that lie within a larger needle trace corresponding to a prostate tissue biopsy (called biopsy core). These ROI-scale models suffer from weak labeling as histopathology results available for biopsy cores only approximate the distribution of cancer in the ROIs. ROI-scale models do not take advantage of contextual information that are normally considered by pathologists, i.e., they do not consider information about surrounding tissue and larger-scale trends when identifying cancer. We aim to improve cancer detection by taking a multi-scale, i.e., ROI-scale and biopsy core-scale, approach. METHODS: Our multi-scale approach combines (i) an "ROI-scale" model trained using self-supervised learning to extract features from small ROIs and (ii) a "core-scale" transformer model that processes a collection of extracted features from multiple ROIs in the needle trace region to predict the tissue type of the corresponding core. Attention maps, as a by-product, allow us to localize cancer at the ROI scale. RESULTS: We analyze this method using a dataset of micro-ultrasound acquired from 578 patients who underwent prostate biopsy, and compare our model to baseline models and other large-scale studies in the literature. Our model shows consistent and substantial performance improvements compared to ROI-scale-only models. It achieves [Formula: see text] AUROC, a statistically significant improvement over ROI-scale classification. We also compare our method to large studies on prostate cancer detection, using other imaging modalities. CONCLUSIONS: Taking a multi-scale approach that leverages contextual information improves prostate cancer detection compared to ROI-scale-only models. The proposed model achieves a statistically significant improvement in performance and outperforms other large-scale studies in the literature. Our code is publicly available at www.github.com/med-i-lab/TRUSFormer .

Non-parametric combination of multimodal MRI for lesion detection in focal epilepsy.

One third of patients with medically refractory focal epilepsy have normal-appearing MRI scans. This poses a problem as identification of the epileptogenic region is required for surgical treatment. This study performs a multimodal voxel-based analysis (VBA) to identify brain abnormalities in MRI-negative focal epilepsy. Data was collected from 69 focal epilepsy patients (42 with discrete lesions on MRI scans, 27 with no visible findings on scans), and 62 healthy controls. MR images comprised T1-weighted, fluid-attenuated inversion recovery (FLAIR), fractional anisotropy (FA) and mean diffusivity (MD) from diffusion tensor imaging, and neurite density index (NDI) from neurite orientation dispersion and density imaging. These multimodal images were coregistered to T1-weighted scans, normalized to a standard space, and smoothed with 8 mm FWHM. Initial analysis performed voxel-wise one-tailed t-tests separately on grey matter concentration (GMC), FLAIR, FA, MD, and NDI, comparing patients with epilepsy to controls. A multimodal non-parametric combination (NPC) analysis was also performed simultaneously on FLAIR, FA, MD, and NDI. Resulting p-maps were family-wise error rate corrected, threshold-free cluster enhanced, and thresholded at p < 0.05. Sensitivity was established through visual comparison of results to manually drawn lesion masks or seizure onset zone (SOZ) from stereoelectroencephalography. A leave-one-out cross-validation with the same analysis protocols was performed on controls to determine specificity. NDI was the best performing individual modality, detecting focal abnormalities in 38% of patients with normal MRI and conclusive SOZ. GMC demonstrated the lowest sensitivity at 19%. NPC provided superior performance to univariate analyses with 50% sensitivity. Specificity in controls ranged between 96 and 100% for all analyses. This study demonstrated the utility of a multimodal VBA utilizing NPC for detecting epileptogenic lesions in MRI-negative focal epilepsy. Future work will apply this approach to datasets from other centres and will experiment with different combinations of MR sequences.