In this work, we propose a smartphone-based system for detecting diabetic foot ulcers. Existing solutions mostly rely on clinical tests or costly imaging systems such as MRI or X-rays. These face challenges in resource-constrained environments and also do not provide immediate results. To address these limitations, we propose IgniSole, an edge-based offline solution for the detection of diabetic foot ulcers. We first train a fine-tuned MobileNetV2 Convolutional Neural Network (CNN) on the plantar thermogram dataset using transfer learning. We also develop an Android application to allow users to test their diabetic foot images using a smartphone. We deploy the trained model on the Android application to reduce dependency on the internet and to secure the user’s data. We evaluate the model’s performance in two phases: the training phase and the inference phase on the smartphone. We observe that our model achieves an accuracy of 96.24%, specificity of 100%, and sensitivity of 92.85% in the training phase. In contrast, the model’s accuracy, specificity, and sensitivity are 96.16%, 98.88%, and 97.96%, respectively, in the inference phase on the smartphone. Moreover, the model predicts the results within a span of 1-2 seconds. Therefore, our work provides a smartphone-based platform to users for the detection of diabetic foot ulcers with affordable, scalable, and accurate diagnostic tools. The source codes are available at https://github.com/anshita510/IgniSole.

Congenital heart disease (CHD) is a leading cause of morbidity and mortality in infants and early diagnosis with access to specific diagnostic tools is often limited and the interpretation of heart sounds is subjective. Traditional auscultation relies heavily on clinician skill and is highly variable, while advanced imaging techniques can be expensive, or even impossible in low-resource settings. In this study, we present a multimodal machine-learning framework that combines demographic data and 66 hand-crafted features from denoised and resampled phonocardiogram (PCG) recordings. Signal preprocessing included a third-order Butterworth bandpass filter set between$65-1000 ~\text{Hz}$and resampling to uniform time series recordings of 2000 Hz to preserve signal integrity and reduce noise. For classification we utilize LightGBM achieving binary task accuracy of$94 \%$and multi-class murmur classification accuracy of 87%. The entire system can be deployed to a Raspberry Pi 4 connected to a digital stethoscope and able to conduct real-time inference as an embedded system without cloud or internet access. Experimental results verified high accuracy and low latency operational characteristics making it well suited for an embedded deployment with such limited computation resources. The proposed end-to-end framework provides an affordable, portable, and clinically useful tool to assist in the early detection of abnormal heart sounds and has the potential to revolutionize CHD screening in resource challenged communities.

In the dynamic realm of the internet of vehicles, ensuring robust Vehicle-to-Everything (V2X) communication is essential for advancing intelligent transportation systems. The IEEE 802.11p standard, while foundational, faces challenges in channel estimation due to its limited pilot structure, especially under high-mobility scenarios. Traditional pilot-aided estimation techniques often grapple with error propagation and diminished accuracy in such environments. Addressing these challenges, this paper introduces ‘HAN’, an innovative hybrid attention-based deep learning model that synergizes time-domain and frequency-domain attention mechanisms to enhance channel estimation. Empirical evaluations reveal that HAN achieves up to a 4 dB performance improvement in high signal-to-noise ratio conditions. Comprehensive testing across various channel conditions, modulation schemes, and vehicular speeds within the vehicle-to-vehicle expressway (VTV-EX) channel underscores its resilience and efficacy. Furthermore, implementation on the Xilinx ZCU102 FPGA platform demonstrates the model’s practicality for real-time applications in resource-constrained environments.

Colorectal cancer, a highly lethal form of cancer, can be treated effectively if detected early. However, the current diagnosis process involves a time-consuming and manual review of CT scans to identify cancerous regions and behavior, leading to resource consumption, subjectivity, and dependency on manual assessment. We propose a 3-phase deep neural system for automated colorectal cancer detection using CT scan images to address these challenges. It includes a SegNet network to identify tumor locations, an InceptionResNet V2 network to classify tumors as benign or malignant, and an analysis of tumor area cum perimeter to predict the cancer stage. The proposed model offers a fully automated solution by combining these functionalities under a single umbrella. In real-life CT scans from 37 patients, the proposed model achieved 95.8% ROI segmentation accuracy, a dice coefficient of 0.6214, 69.75% IoU score, and 95.83% tumor classification accuracy. The unique approach using Radial Length (RL) and Circularity (C) parameters predicted the T-stage with close to 85% accuracy. Based on these outcomes, the proposed system establishes itself as a reliable and suitable alternative to traditional cancer diagnosis techniques by leveraging the power of automation, deep learning, and innovative parameter analysis.

LoRa technology endows unprecedented ability to connect isolated geographical landscapes and build community networks that serve specific purposes. As the network grows, coherent routing of messages becomes imperative to meet the network’s objectives and Quality of Service (QoS) requirements. However, despite the recent rise in research and development centered around LoRa networks, not much research addresses the routing mechanisms in LoRa networks. Moreover, the LoRa routing mechanisms must run on low-power and resource-constrained devices, as these networks primarily target far-off locations or volatile environments such as volcanoes. Hence, this work proposes routing mechanisms (LoRaute) for LoRa networks that help route messages considering the messages’ QoS requirements. Also, a multipurpose network hardware is proposed, which serves as a LoRa network base station or a WiFi to LoRa bridge for TCP/IP communication. Additionally, this work furnishes and assesses the implementation of the LoRa network and the routing mechanisms in a real environment. The implementation results indicate a seamless and rapid setup of multihop LoRa networks. Moreover, the implementation achieves a routing table record size of 9 B, 24.45 ms routing latency, and 171 mA peak current consumption by the proposed LoRa node. Finally, the proposed system serves as a backhaul network for essential long-range communications, as demonstrated by the experimental setup.

In social IoMT systems, resource-constrained devices face the challenges of limited computation, bandwidth, and privacy in the deployment of deep learning models. Federated learning (FL) is one of the solutions to user privacy and provides distributed training among several local devices. In addition, it reduces the computation and bandwidth of transferring videos to the central server in camera-based IoMT devices. In this work, we design an edge-based federated framework for such devices. In contrast to traditional methods that drop the resource-constrained stragglers in a federated round, our system provides a methodology to incorporate them. We propose a new phase in the FL algorithm, known as split learning. The stragglers train collaboratively with the nearest edge node using split learning. We test the implementation using heterogeneous computing devices that extract vital signs from videos. The results show a reduction of 3.6 h in the training time of videos using the split learning phase with respect to the traditional approach. We also evaluate the performance of the devices and system with key parameters, CPU utilization, memory consumption, and data rate. Furthermore, we achieve 87.29% and 60.26% test accuracy at the nonstragglers and stragglers, respectively, with a global accuracy of 90.32% at the server. Therefore, FedCare provides a straggler-resistant federated method for a heterogeneous system for social IoMT devices.

The social IoMT-based activity-monitoring system comprises several devices with different datasets. It faces challenges like a collection of a global activity dataset which comprises a myriad of activities. In this article, we propose a federated Siamese network-based data-independent group activity segregator—Skipper—which aims to identify anomalies in an activity-monitoring social IoMT system. The novelty of this work is that Skipper does not require any dataset before its deployment, which removes the need for any prior training of the model for activity monitoring. As a proof of concept, we select activities pertaining to school environments to identify low-performing students in a classroom, who would require teachers’ close attention to ensure balanced growth and proper health. Skipper monitors the students independently for their motion signatures through a wearable device that consists of an accelerometer. A federated Siamese network calculates indices that signify the degree of similarity among the students’ activities. Skipper identifies the students who do not perform the same activity. With real-world implementations, we observe that Skipper requires network rates of 10 Kb/s, making it suitable for low bandwidth networks while we achieve just 20% CPU and 10 MB memory utilization on constrained edge devices. Further, with an increasing number of students up to 100, the time delay for final results is limited to 80 s. Hence, Skipper is a fast, easy, and accurate solution for recognizing outliers in IoMT social systems.

Detection of COVID-19 has been a global challenge due to the lack of proper resources across all regions. Recently, research has been conducted for non-invasive testing of COVID-19 using an individual's cough audio as input to deep learning models. However, these methods do not pay sufficient attention to resource and infrastructure constraints for real-life practical deployment and the lack of focus on maintaining user data privacy makes these solutions unsuitable for large-scale use. We propose a resource-efficient CoviFL framework using an AIoMT approach for remote COVID-19 detection while maintaining user data privacy. Federated learning has been used to decentralize the CoviFL CNN model training and test the COVID-19 status of users with an accuracy of 93.01 % on portable AIoMT edge devices. Experiments on real-world datasets suggest that the proposed CoviF L solution is promising for large-scale deployment even in resource and infrastructure-constrained environments making it suitable for remote COVID-19 detection.

The paper presents a novel, self-sufficient, Internet of Medical Things-based model called iNAP to address the shortcomings of anemia and polycythemia detection. The proposed model captures eye and fingernail images using a smartphone camera and automatically extracts the conjunctiva and fingernails as the regions of interest. A novel algorithm extracts the dominant color by analyzing color spectroscopy of the extracted portions and accurately predicts blood hemoglobin level. A less than 11.5 gdL \( ^{-1} \) value is categorized as anemia while a greater than 16.5 gdL \( ^{-1} \) value as polycythemia. The model incorporates machine learning and image processing techniques allowing easy smartphone implementation. The model predicts blood hemoglobin to an accuracy of \( \pm \) 0.33 gdL \( ^{-1} \) , a bias of 0.2 gdL \( ^{-1} \) , and a sensitivity of 90 \( \% \) compared to clinically tested results on 99 participants. Furthermore, a novel brightness adjustment algorithm is developed, allowing robustness to a wide illumination range and the type of device used. The proposed IoMT framework allows virtual consultations between physicians and patients, as well as provides overall public health information. The model thereby establishes itself as an authentic and acceptable replacement for invasive and clinically-based hemoglobin tests by leveraging the feature of self-anemia and polycythemia diagnosis.

In this work, a neural network (NN)-based detection for mobile molecular communication via diffusion (MCvD) is proposed. The proposed detector employs a scaled conjugate gradient (SCG) algorithm for updating the weights of the NN. Moreover, three different techniques are used in training and detection by the NN. These techniques correspond to i) filtered signal, ii) slope values of the filtered signal, and iii) concentration difference of the filtered signal in a bit interval. More specifically, a sequence of transmitted bit pattern and each of the above three techniques are used separately to train the NN. After training, the NN-based detector performs detection under a time-varying channel. The bit error rate (BER) performance of the proposed SCG algorithm for the NN-based detector is also compared with a first-order algorithm Gradient Descent (GD) and a second-order algorithm Broyden-Fletcher-Goldfarb-Shanno (BFGS) for different coherence times of the channel. Simulation results demonstrate that the NN detector using SCG outperforms the BFGS if slope values are used for training the NN. Further, the SCG algorithm has a significant performance gain compared to the GD algorithm.

Due to the dynamic nature of nanomachines and diffusion channel, detection is challenging in mobile molecular communication (MMC). In general, signal detection is performed on the number of received molecules with respect to a fixed threshold. However, in MMC number of received molecules varies in each bit interval, which makes detection very challenging if a fixed threshold is considered. In addition to this, the dynamic nature of nanomachines, communication channel impacted inter-symbol interference (ISI), and makes it time-varying. In this work, we propose a detection technique with adaptive threshold for signal detection in MMC. The selection of adaptive threshold depends on the parameters such as the dynamic distance between nanomachines, diffusion coefficient etc. Detection using adaptive threshold improves the detection performance and also mitigates ISI manifold. In this paper, we use modified concentration shift keying (M-CSK) based modulation to release more amount of signaling molecules for bit-1 than bit-0. Received signal is reconstructed in each bit interval using the average distance variation in a bit interval and the reconstructed received signal is subtracted from the total received signal in subsequent bit duration. Detection threshold is calculated by applying maximum a posteriori (MAP) rule to reconstructed signals for the bit-1 and the bit-0 of previous bit interval. Performance of this detector is compared with an existing detection technique. Results reveal a better bit error rate (BER) performance in terms of parameters like coherence time of the channel (zero BER for coherence time of two bit durations), increasing diffusion coefficient and initial distance between nanomachines.