3DKnITS: Three-dimensional Digital Knitting of Intelligent Textile Sensor for Activity Recognition and Biomechanical Monitoring.

We present an approach to develop seamless and scalable piezo-resistive matrix-based intelligent textile using digital flat-bed and circular knitting machines. By combining and customizing functional and common yarns, we can design the aesthetics and architecture and engineer both the electrical and mechanical properties of a sensing textile. By incorporating a melting fiber, we propose a method to shape and personalize three-dimensional piezo-resistive fabric structure that can conform to the human body through thermoforming principles. It results in a robust textile structure and intimate interfacing, suppressing sensor drifts and maximizing accuracy while ensuring comfortability. This paper describes our textile design, fabrication approach, wireless hardware system, deep-learning enabled recognition methods, experimental results, and application scenarios. The digital knitting approach enables the fabrication of 2D to 3D pressure-sensitive textile interiors and wearables, including a 45 x 45 cm intelligent mat with 256 pressure-sensing pixels, and a circularly-knitted, form-fitted shoe with 96 sensing pixels across its 3D surface both with linear piezo-resistive sensitivity of 39.4 for up to 500 N load. Our personalized convolutional neural network models are able to classify 7 basic activities and exercises and 7 yoga poses in-real time with 99.6% and 98.7% accuracy respectively. Further, we demonstrate our technology for a variety of applications ranging from rehabilitation and sport science, to wearables and gaming interfaces.

Pan-cancer Transcriptomic Predictors of Perineural Invasion Improve Occult Histopathologic Detection.

PURPOSE: Perineural invasion (PNI) is associated with aggressive tumor behavior, recurrence, and metastasis, and can influence the administration of adjuvant treatment. However, standard histopathologic examination has limited sensitivity in detecting PNI and does not provide insights into its mechanistic underpinnings. EXPERIMENTAL DESIGN: A multivariate Cox regression was performed to validate associations between PNI and survival in 2,029 patients across 12 cancer types. Differential expression and gene set enrichment analysis were used to learn PNI-associated programs. Machine learning models were applied to build a PNI gene expression classifier. A blinded re-review of hematoxylin and eosin (H&E) slides by a board-certified pathologist helped determine whether the classifier could improve occult histopathologic detection of PNI. RESULTS: PNI associated with both poor overall survival [HR, 1.73; 95% confidence interval (CI), 1.27-2.36; P < 0.001] and disease-free survival (HR, 1.79; 95% CI, 1.38-2.32; P < 0.001). Neural-like, prosurvival, and invasive programs were enriched in PNI-positive tumors (P adj < 0.001). Although PNI-associated features likely reflect in part the increased presence of nerves, many differentially expressed genes mapped specifically to malignant cells from single-cell atlases. A PNI gene expression classifier was derived using random forest and evaluated as a tool for occult histopathologic detection. On a blinded H&E re-review of sections initially described as PNI negative, more specimens were reannotated as PNI positive in the high classifier score cohort compared with the low-scoring cohort (P = 0.03, Fisher exact test). CONCLUSIONS: This study provides salient biological insights regarding PNI and demonstrates a role for gene expression classifiers to augment detection of histopathologic features.