Cancer is the second leading cause of death globally, and breast cancer (BC) is the second most reported cancer. Although the incidence rate is reducing in developed countries, the reverse is the case in low- and middle-income countries. Early detection has been found to contain cancer growth, prevent metastasis, ease treatment, and reduce mortality by 25%. The digital mammogram is one of the most common, cheapest, and most effective BC screening techniques capable of early detection of up to 90% BC incidence. However, the mammogram is one of the most difficult medical images to analyze. In this paper, we present a method of training a deep learning model for BC diagnosis. We developed a discriminative fine-tuning method which dynamically assigns different learning rates to each layer of the deep CNN. In addition, the model was trained using mixed-precision training to ease the computational demand of training deep learning models. Lastly, we present data augmentation methods for mammograms. The discriminative fine-tuning algorithm enables rapid convergence of the model loss; hence, the models were trained to attain their best performance within 50 epochs. Comparing the results, DenseNet achieved the highest accuracy of 0.998, while AlexNet obtained 0.988.
The first and most critical response to curbing the spread of the novel coronavirus disease (COVID-19) is to deploy effective techniques to test potentially infected patients, isolate them and commence immediate treatment. However, several test kits currently in use are slow and in a shortage of supply. This paper presents techniques for diagnosing COVID-19 from chest X-ray (CXR) and address problems associated with training deep models with less voluminous datasets and class imbalance as obtained in most available CXR datasets on COVID-19. We introduced the discriminative fine-tuning approach, which dynamically assigns different learning rates to each layer of the network. The learning rate is set using the cyclical learning rate policy that changes per iteration. This flexibility ensured rapid convergence and avoided being stuck in saddle point plateau. In addition, we addressed the high computational demand of deep models by implementing our algorithm using the memory- and computational-efficient mixed-precision training. Despite the availability of scanty datasets, our model achieved high performance and generalisation. A Validation accuracy of 96.83%, sensitivity and specificity of 96.26% and 95.54% were obtained, respectively. When tested on an entirely new dataset, the model achieves 97% accuracy without further training. Lastly, we presented a visual interpretation of the model’s output to prove that the model can aid radiologists in rapidly screening for the symptoms of COVID-19.
This work addresses the problem of intra-class classification of Breast Histopathology images into Eight (8) classes of either Benign or Malignant Cell. Current manual features extraction and classification is fraught with inaccuracies leading to high rate false negatives with attendant mortality. Deep Convolutional Neural Networks (DCNN) have been shown to be effective in classification of Images. We adopted a DCNN architecture combined with Ensemble learning method using TensorFlow Framework with Backpropagation training and ReLU activation function to achieve accurate automated classification of these Images. We achieved inter-class classification accuracy of 91.5% with the BreakHis dataset.
1 April 2017 in Engineering Science and Technology, an International Journal
The possibility of seamless handover remains a mirage despite the plethora of existing handover algorithms. The underlying factor responsible for this has been traced to the Handover decision module in the Handover process. Hence, in this paper, the development of novel hybrid artificial intelligent handover decision algorithm has been developed. The developed model is made up of hybrid of Artificial Neural Network (ANN) based prediction model and Fuzzy Logic. On accessing the network, the Received Signal Strength (RSS) was acquired over a period of time to form a time series data. The data was then fed to the newly proposed k-step ahead ANN-based RSS prediction system for estimation of prediction model coefficients. The synaptic weights and adaptive coefficients of the trained ANN was then used to compute the k-step ahead ANN based RSS prediction model coefficients. The predicted RSS value was later codified as Fuzzy sets and in conjunction with other measured network parameters were fed into the Fuzzy logic controller in order to finalize handover decision process. The performance of the newly developed k-step ahead ANN based RSS prediction algorithm was evaluated using simulated and real data acquired from available mobile communication networks. Results obtained in both cases shows that the proposed algorithm is capable of predicting ahead the RSS value to about ±0.0002 dB. Also, the cascaded effect of the complete handover decision module was also evaluated. Results obtained show that the newly proposed hybrid approach was able to reduce ping-pong effect associated with other handover techniques
Breast Cancer is one of the most diagnosed cancer and the leading cause of death among women worldwide, second only to lung cancer. Mammographic screening has been the most successful screening technology capable of detecting up to 90% of all breast cancer even before a lump growth can be felt using breast exam. However, mammogram is a low intensity image and the heterogeneous nature of breast can make healthy breast tissue appears as cancerous, this is most common among women with dense breast (aged 40-44). Thus, the sensitivity for early detection of breast cancer from mammogram has been estimated to 85-90%. This result can be improved by Deep CNN, however, to achieve good generalization, it must be train with high voluminous dataset whereas, mammographic dataset exists in smaller volume. In this paper, we present a method of training deep CNN with few datasets to achieve high training result and good generalization. An augmentation technique that increase both size and variance of the dataset is presented herewith, the augmented dataset was used to train five state of the art models. Highest training and validation accuracy (99.01% and 99.99% respectively) were achieved with DensNet. Meanwhile, SqueezeNet, a deep CNN model with fewer parameter also shows promising result, which means soon this model can be deployed into microcontroller and FPGAs for clinical applications.
This paper focuses on the development of maize plant datasets for the purposes of recognizing maize plants and weed species, as well as the precise automated application of herbicides to the weeds. The dataset includes 36,374 images captured with a high-resolution digital camera during the weed survey and 500 images annotated with the Labelmg suite. Images of the eighteen farmland locations in North Central Nigeria, containing the maize plants and their associated weeds were captured using a high-resolution camera in each location. This dataset will serve as a benchmark for computer vision and machine learning tasks in the intelligent maize and weed recognition research.