Pneumonia classification: A limited data approach for global understanding.

As the human race has advanced, so too have the ailments that afflict it. Diseases such as pneumonia, once considered to be basic flu or allergies, have evolved into more severe forms, including SARs and COVID-19, presenting significant risks to people worldwide. In our study, we focused on categorizing pneumonia-related inflammation in chest X-rays (CXR) using a relatively small dataset. Our approach was to encompass a comprehensive view, addressing every potential area of inflammation in the CXR. We employed enhanced class activation maps (mCAM) to meet the clinical criteria for classification rationale. Our model incorporates capsule network clusters (CNsC), which aids in learning different aspects such as geometry, orientation, and position of the inflammation seen in the CXR. Our Capsule Network Clusters (CNsC) rapidly interpret various perspectives in a single CXR without needing image augmentation, a common necessity in existing detection models. This approach significantly cuts down on training and evaluation durations. We conducted thorough testing using the RSNA pneumonia dataset of CXR images, achieving accuracy and recall rates as high as 98.3% and 99.5% in our conclusive tests. Additionally, we observed encouraging outcomes when applying our trained model to standard X-ray images obtained from medical clinics.

A Preemptive Priority-Based Data Fragmentation Scheme for Heterogeneous Traffic in Wireless Sensor Networks.

Prioritizing the heterogeneous traffic for Wireless Sensor Networks (WSNs) imposes an important performance challenge for Internet of Things (IoT) applications. Most past preemptive MAC schemes are based on scheduling the high priority packets earlier than those of lower priority. However, in a majority of these schemes, high priority traffic must wait for the ongoing transmission of lower priority traffic due to the non-availability of an interruption mechanism. This paper presents the design and high-level implementation details of a fragmentation scheme (FROG-MAC) for heterogeneous traffic in WSN. FROG-MAC aims at guaranteeing quick transmission of high priority/emergency traffic by interrupting ongoing on channel transmissions. High level implementation of FROG-MAC has been developed in MATLAB as a proof of concept. Traffic of two priorities was generated and a single hop star topology of 100 nodes was used for the experiments. Effect of the proposed fragmentation scheme has been evaluated on delay and Packet Drop Ratio (PDR) for both traffic types, by varying the packet size and fragment size. Simulation results have suggested that with the increasing packet size, the delay and PDR increase for both traffic types. When fragmentation was applied, the performance of high priority traffic significantly improved as compared to the low priority for both the parameters, delay and PDR. Furthermore, it has been found that decreasing the fragment size for low priority traffic results in reducing the delay for high priority traffic.

Real-Time Analysis of a Sensor's Data for Automated Decision Making in an IoT-Based Smart Home.

IoT devices frequently generate large volumes of streaming data and in order to take advantage of this data, their temporal patterns must be learned and identified. Streaming data analysis has become popular after being successfully used in many applications including forecasting electricity load, stock market prices, weather conditions, etc. Artificial Neural Networks (ANNs) have been successfully utilized in understanding the embedded interesting patterns/behaviors in the data and forecasting the future values based on it. One such pattern is modelled and learned in the present study to identify the occurrence of a specific pattern in a Water Management System (WMS). This prediction aids in making an automatic decision support system, to switch OFF a hydraulic suction pump at the appropriate time. Three types of ANN, namely Multi-Input Multi-Output (MIMO), Multi-Input Single-Output (MISO), and Recurrent Neural Network (RNN) have been compared, for multi-step-ahead forecasting, on a sensor's streaming data. Experiments have shown that RNN has the best performance among three models and based on its prediction, a system can be implemented to make the best decision with 86% accuracy.