Accurate measurement of the vibrotactile acuities of the human torso is the key to designing effective torso-worn vibrotactile displays for healthcare applications such as navigation aids for visually impaired persons. Although efforts have been made to measure vibrotactile acuities, there remains a lack of systematic studies addressing the spatial, temporal, and intensity-related aspects of vibrotactile sensitivity on the human torso. In this work, a torso-worn vibrotactile belt consisting of two crossed coin motor arrays was designed and a psychophysical study was carried out to measure the spatial, temporal, and intensity-related vibrotactile acuities of a set of human subjects wearing the designed belt. The objective parameters of vibrational intensity and the timing latency of the coin motor were also determined before measuring the vibrotactile acuities. The experimental results indicated that the tested coin motor was able to generate a median number of five and six available just-noticeable differences in intensity and duration, respectively. Among the four parameters of vibrational intensity, the perceived intensity was the most relevant to vibrational displacement. The spatial acuities measured as the degree of two-point spatial thresholds (TPTs) showed less individual difference than the distance TPTs. The results from the current work provide valuable guidance for the design of a comfortable torso-worn vibrotactile display using coin motors.
Lymph node metastasis (LNM) is a key prognostic factor in esophageal squamous cell carcinoma (ESCC), and accurate preoperative prediction remains challenging. Blood biomarkers provide a conventional, preoperative diagnostic technique that is cost-effective and free from radiation risks. So far, previous studies have been published on the precise diagnosis of lymph node metastases using conventional ultrasound or CT techniques. While there is a lack of research studies that address the diagnosis of LNM from blood biomarkers. In this work, we acquired a cohort of blood biomarkers of 1933 patients and designed a weighted neural network (WNN) model for the accurate prediction of LNM from blood biomarkers. The WNN model is designed with a neural network classifier trained on blood biomarkers labeled with pathological nodal (pN) stages of LNM. The experimental findings demonstrate that the WNN model achieved 83.1% accuracy and an AUC of 0.88 on the original, non-augmented test set for diagnosing LNM, while CT only achieved 50.4% accuracy (AUC 0.60) and ultrasound achieved 60.5% accuracy (AUC 0.67). Additionally, SHAP analysis reveals that three blood biomarkers—white blood cells (WBC#), monocytes (Mono#), and neutrophils (Neut#)—significantly impact the WNN model’s output. This WNN model shows promise as a research tool to diagnose LNM.