Duygu Sinanc Terzi’s research while affiliated with Amasya University and other places

The weight initialization technique for transfer learning refers to the practice of using pretrained models that can be modified to solve new problems, instead of starting the training process from scratch. In this study, six different transfer learning weight initialization strategies were proposed for plant disease detection: scratch (i.e., random initialization), pretrained model on cross‐domain (ImageNet), model trained on related domain (ISIC 2019), model trained on related domain (ISIC 2019) with cross‐domain (ImageNet) weights, model trained on same domain (PlantVillage), and model trained on same domain (PlantVillage) with cross‐domain weights (ImageNet). Weights from each strategy were transferred to a target dataset (Plant Pathology 2021). These strategies were implemented using eight deep learning architectures. It was observed that transferring from any strategy led to an average acceleration of convergence ranging from 33.88% to 73.16% in mean loss and an improvement of 8.72%–42.12% in mean F 1 ‐score compared to the scratch strategy. Moreover, although smaller and less comprehensive than ImageNet, transferring information from the same domain or related domain proved to be competitive compared to transferring from ImageNet. This indicates that ImageNet, which is widely favoured in the literature, may not necessarily represent the most optimal transfer source for the given context. In addition, to identify which strategies have significant differences, a post hoc analysis using Tukey's HSD test was conducted. Finally, the classifications made by the proposed models were visualized using Grad‐CAM to provide a qualitative understanding of how different weight initialization strategies affect the focus areas of the models.

The concept of weight initialization technique for transfer learning refers to the practice of using pre-trained models that can be modified to solve new problems, instead of starting the training process from scratch. By using pre-trained models as a starting point, the network can learn from patterns and features present in the original data, improving overall accuracy and allowing for faster convergence during training. In this study, four different transfer learning weight initialization strategies are proposed for plant disease detection: random initialization, pre-trained model on different domain (ImageNet), model trained on related domain (ISIC 2019), and model trained on same domain (PlantVillage). Weights from each strategy are transferred to a target dataset, Plant Pathology 2021. These strategies were implemented using four state-of-the-art CNN-based architectures: AlexNet, DenseNet, MobileNetV2, and VGG. The best result was obtained when both the target and source datasets included images of plant diseases. In this case, VGG was used and resulted in an 85.9% weighted f-score, which is a 9% improvement from random initialization. The transfer of knowledge from small-sized, related domain data (skin cancer data) was almost as successful as the transfer from ImageNet. Transferring from ImageNet yielded an f-score of 85.7%, while transferring from skin cancer data resulted in an f-score of 85.2%. This indicates that ImageNet, which is widely favored in the literature, may not necessarily represent the most optimal transfer source for the given context. Finally, the classifications made by the proposed models were visualized using Grad-CAM to better understand the decision-making process.

Transfer learning has gained importance in areas where there is a labeled data shortage. However, it is still controversial as to what extent natural image datasets as pre-training sources contribute scientifically to success in different fields, such as medical imaging. In this study, the effect of transfer learning for medical object detection was quantitatively compared using natural and medical image datasets. Within the scope of this study, transfer learning strategies based on five different weight initialization methods were discussed. A natural image dataset MS COCO and brain tumor dataset BraTS 2020 were used as the transfer learning source, and Gazi Brains 2020 was used for the target. Mask R-CNN was adopted as a deep learning architecture for its capability to effectively handle both object detection and segmentation tasks. The experimental results show that transfer learning from the medical image dataset was found to be 10% more successful and showed 24% better convergence performance than the MS COCO pre-trained model, although it contains fewer data. While the effect of data augmentation on the natural image pre-trained model was 5%, the same domain pre-trained model was measured as 2%. According to the most widely used object detection metric, transfer learning strategies using MS COCO weights and random weights showed the same object detection performance as data augmentation. The performance of the most effective strategies identified in the Mask R-CNN model was also tested with YOLOv8. Results showed that even if the amount of data is less than the natural dataset, in-domain transfer learning is more efficient than cross-domain transfer learning. Moreover, this study demonstrates the first use of the Gazi Brains 2020 dataset, which was generated to address the lack of labeled and qualified brain MRI data in the medical field for in-domain transfer learning. Thus, knowledge transfer was carried out from the deep neural network, which was trained with brain tumor data and tested on a different brain tumor dataset.

Cyber threats are increasing progressively in their frequency, scale, sophistication, and cost. The advancement of such threats has raised the need to enhance intelligent intrusion-detection systems. In this study, a different perspective has been developed for intrusion detection. Gramian angular fields were adapted to encode network traffic data as images. Hereby, a way to reveal bilateral feature relationships and benefit from the visual interpretation capability of deep-learning methods has been opened. Then, image-encoded intrusions were classified as binary and multi-class using convolutional neural networks. The obtained results were compared to both conventional machine-learning methods and related studies. According to the results, the proposed approach surpassed the success of traditional methods and produced success rates that were close to the related studies. Despite the use of complex mechanisms such as feature extraction, feature selection, class balancing, virtual data generation, or ensemble classifiers in related studies, the proposed approach is fairly plain -- involving only data-image conversion and classification. This shows the power of simply changing the problem space.

This survey aims to identify commonly used methods, datasets, future trends, knowledge gaps, constraints, and limitations in the field to provide an overview of current solutions used in medical image analysis in parallel with the rapid developments in transfer learning (TL). Unlike previous studies, this survey grouped the last five years of current studies for the period between January 2017 and February 2021 according to different anatomical regions and detailed the modality, medical task, TL method, source data, target data, and public or private datasets used in medical imaging. Also, it provides readers with detailed information on technical challenges, opportunities, and future research trends. In this way, an overview of recent developments is provided to help researchers to select the most effective and efficient methods and access widely used and publicly available medical datasets, research gaps, and limitations of the available literature.

One of the problems that often arise during the application of medical research to real life is the high number of false positive cases. This situation causes experts to be warned with false alarms unnecessarily and increases their workload. This study proposes a new data centric approach to reduce bias‐based false positive predictions in brain MRI‐specific medical object detection applications. The proposed method has been tested using two different datasets: Gazi Brains 2020 and BraTS 2020, and three different deep learning‐based object detection models: Mask R‐CNN, YOLOv5, and EfficientDet. According to the results, the proposed pipeline outperformed the classical pipeline, up to 18% on the Gazi Brains 2020 dataset, and up to 24% on the BraTS 2020 dataset for mean specificity value without much change in sensitivity metric. It means that the proposed pipeline reduces false positive rates due to bias in real‐life applications and it can help to reduce the workload of experts.

Fraud detection; Time series; Deep learning; Explainable artificial intelligence; Image conversion The increase in the volume and velocity of credit card transactions causes class imbalance and concept deviation problems in data sets where credit card fraud is detected. These problems make it very difficult for traditional approaches to produce robust detection models. In this study, a different perspective has been developed for this problem and a novel approach named Fraud Detection with Image Conversion (FDIC) is proposed. FDIC handles credit card transactions as time series and transforms them into images. These images, which comprise temporal correlations and bilateral relationships of features, are classified by a convolutional neural network architecture as fraudulent or legitimate. When the obtained results are compared with the related studies, FDIC has the best F1-score and recall values, which are 85.49% and 80.35%, respectively. This shows that FDIC is better than other studies in detecting fraudulent instances associated with high cost. Since the images created during the FDIC process are difficult to interpret, a new explainable artificial intelligence approach is also presented. In this way, feature relationships that have a dominant effect on fraud detection are revealed.

The class imbalance problem, one of the common data irregularities, causes the development of under-represented models. To resolve this issue, the present study proposes a new cluster-based MapReduce design, entitled Distributed Cluster-based Resampling for Imbalanced Big Data (DIBID). The design aims at modifying the existing dataset to increase the classification success. Within the study, DIBID has been implemented on public datasets under two strategies. The first strategy has been designed to present the success of the model on data sets with different imbalanced ratios. The second strategy has been designed to compare the success of the model with other imbalanced big data solutions in the literature. According to the results, DIBID outperformed other imbalanced big data solutions in the literature and increased area under the curve values between 10 % and 24 % through the case study.