Mücella Özbay Karakuş

Classification of different types of strawberries is an essential problem in agriculture and quality control; however, it presents a challenge to traditional deep learning solutions because of the need for large-scale labeled datasets. In this research, a few-shot learning (FSL) model based on Prototypical Networks was applied to classify four different commercial types of strawberries (A, B, C, and Jumbo), utilizing the Mendeley Strawberry Grading Dataset that includes only 103 labeled images for training. A unique pipeline for training an FSL model was created through a combination of warm-up pre-training, differential learning rate strategy, and 4-way 5-shot episodic training; two backbones (ResNet50 and EfficientNet-B0) were compared under a 5-fold stratified cross-validation. While ResNet50 obtained test accuracy, macro F1-score, and Cohen's Kappa equal to 99.00%, 99.11%, and 98.62% respectively, EfficientNet-B0 achieved 98.00%, 98.33%, and 97.30%. Both models produced an AUC value of 99.80%. Interpretability analysis utilizing GradCAM++ showed that both networks focused on relevant areas of the fruits. This research proved that metric-based few-shot learning can be used successfully as an alternative solution to traditional deep learning methods in case of scarce datasets of agricultural images.

Breast cancer remains a serious global health threat, and there is a clear need for diagnostically reliable research frameworks that go beyond the overly optimistic performance claims often reported in the literature. This study provides a clinical diagnostic methodology for breast cancer using fine-needle aspiration cytology, which is intended to promote clinical validation beyond benchmark maximization. The methodology presented uses Wisconsin Diagnostic Breast Cancer and Wisconsin Prognostic Breast Cancer datasets to develop a completely leak-proof nested 5x5 cross-validation procedure where any preprocessing was limited to training folds only. The original 30 cytology features were increased to 42 by adding biologically-motivated feature engineering, and a stacked ensemble method was trained to provide clinically interpretable probability estimates. The method used bootstrap-based feature stability, split conformal prediction to perform uncertainty-driven triage, feature attribution and counterfactual explanation modules, decision curve analysis for clinical validation, and multi-task learning for cross-dataset transfer analysis. In addition, the method yielded an out-of-fold area under the curve of 0.9932 and a holdout accuracy of 0.9912. Given 95% confidence interval, split conformal prediction resulted in a definitive single-class prediction for 99.1% of instances, suggesting that uncertain cases constitute only a negligible number and require only an escalation in clinical practice. The results of decision curve analysis have shown that there is always net clinical benefit across all clinically relevant probability thresholds. Finally, multi-task analysis has shown that cytological morphology is useful for the purposes of breast cancer diagnosis but not prognosis.

Background/Objectives: This paper aims to design an explainable and accurate ML framework to support the automatic classification of Carpal Tunnel Syndrome (CTS) severity from structured patient data. Methods: For the experiment, an open-source dataset of 1037 samples was used. Following stratified partitioning, 305 samples were held out as the test set; the remaining training set (n = 732) was augmented to 1216 balanced samples via ADASYN, yielding an 80/20 train/test ratio relative to the final dataset (n = 1521). In order to solve the problem of imbalance associated with CTS cases of moderate and severe severity, the Adaptive Synthetic Sampling (ADASYN) technique was employed. The model’s predictive capacity was increased by means of feature engineering methods, such as polynomial transformations and clinically relevant interactions. Specifically, four ensemble learning models (XGBoost, Random Forest, LightGBM, and CatBoost) were optimized and ensembled with the use of a stacking approach with a base algorithm of LightGBM. The explainability of the model was ensured through SHAP and LIME analysis. Results: As a result, the stacking ensemble was able to reach a test accuracy of 91.15%, an F1-score of 91.13%, and an ROC-AUC of 0.9708. The proposed ensemble performed superiorly compared to any other individual algorithm while having stable performance across all severity categories. Conclusions: Through the explainability analysis, it was observed that such a classification model relies on important clinically relevant predictors, including cross-sectional area (CSA), duration of symptoms, pain level measured by the numeric rating scale of pain (NRS), and palmar bowing (PB).

This study presents a novel and sustainable approach to dye-sensitized solar cell (DSSC) design by integrating a natural dye extracted from Peltophorum pterocarpum with CdS co-sensitization and ZnO nanorod-modified TiO₂ photoelectrodes. The fabricated DSSCs were enhanced through the integration of ZnO nanorods on TiO₂ photoelectrodes and further modified with CdS co-sensitization to boost their photovoltaic and photodetection capabilities. The devices were evaluated under a broad range of light intensities (100 to 0.01 mW/cm²), revealing dual functionality: efficient photovoltaic operation under standard illumination and superior sensitivity under ultra-low light. Under standard illumination, the highest efficiency (8.15%) was achieved with N719 dye and ZnO/TiO₂ electrodes. However, the P. pterocarpum-based cells also exhibited competitive performance, reaching 6.92% with CdS and ZnO enhancement. Remarkably, under ultra-low light (0.01 mW/cm²), the natural dye-based DSSCs demonstrated power conversion efficiencies exceeding 18%, rivaling commercial dye systems. These results highlight the synergistic role of ZnO nanorods and CdS in enhancing charge transport and light absorption, while also showcasing the viability of natural dyes for eco-friendly, low-cost optoelectronic applications. This study is among the first to demonstrate the synergistic integration of P. pterocarpum natural dye, CdS co-sensitization, and ZnO nanorods in DSSCs and in this research, sustainable natural dye utilization with structural nanomaterial engineering is uniquely bridged to develop DSSCs capable of dual functionality under both standard and ultra-low light conditions.

This study focuses on the effect of cross‐linking mechanisms in the development of hydrogel‐based flexible ionic diodes (HBIDs) with desirable features for ionic devices. The research explores one p‐type hydrogel and a series of n‐type hydrogels with varying cross‐linking ratios, investigating their ion‐conducting properties, morphological and electronic structures, as well as the performance of flexible HBIDs constructed using these hydrogels. The results indicate that the cross‐linking ratio significantly influences ionic conductivity, bulk resistance, and capacitance. Additionally, the HBIDs' ability to maintain their electrical performance under repeated mechanical bending is studied concerning the cross‐linking ratio. Notably, the HBID with the highest cross‐linking ratio (HG/12) exhibits relatively higher current densities (up to 8.9 mA cm⁻²) and considerably higher rectification ratios (up to 1269). Even after 500 cycles of deformation, the HG/12 diode maintained a rectification ratio of 987, demonstrating its excellent performance and durability.

The integration of robotics into various aspects of human life has prompted an urgent need for robots to possess essential skills such as object recognition and navigational proficiency. Recent strides in robotics technology have ushered robots into domains like surveillance, quality control, and assisted living. This study delves into the optimization of neural networks for enabling robot navigation tasks. The investigation employs a support vector machine (SWM) structure and compares its performance against other neural network architectures using the same dataset derived from ultrasound sensor measurements. The SWM exhibits superior accuracy, achieving a remarkable average classification accuracy of 99.93%, outperforming other architectures. The findings underscore the potential of neural networks, particularly the PNN, in advancing robot navigation and fostering their integration into various applications. The study contributes to the growing field of robotics by offering insights into the effectiveness of different neural network models for enhancing mobile robot navigation capabilities.

Beyaz kan hücreleri, insanların bağışıklık sisteminin en önemli yapısı olup, kan ve lenf dokularında kemik iliği tarafından üretilmektedir. Bu hücreler insan vücudunu hastalık ve yabancı organizmalara karşı koruyan savunma mekanizmalarıdır ve kandaki oranı düştüğünde Lökopeni ile karşılaşılabilir. Bu hücrelerin insan vücudundaki oranının belirlenmesi ve hastalığın tespit ve tedavisi için yoğun emek harcaması gerekmektedir. Bu çalışmada, derin öğrenme modellerini kullanarak beyaz kan hücreleri sınıflandırma performansının iyileştirilmesi amaçlanmıştır. Sınıflandırma işlemini daha verimli gerçekleştirmek için evrişimli sinir ağı modelleri kullanılmıştır. Beyaz kan hücresi çeşitleri olan eozinofil, lenfosit, monosit ve nötrofil arasında ayrım yapmak için Densenet201, ResNet50 ve Alexnet birleştirilmiştir. Elde edilen özellik haritalarının sınıflandırılması için K-En yakın komşuluk, Destek Vektör Makinesi ve Naïve Bayes olmak üzere üç farklı makine öğrenmesi sınıflandırıcısı kullanılmıştır. Derin Öğrenme (DÖ) ile eğitilen Kaggle veri kümesi görüntülerine CLAHE ve Gauss filtreleri uygulanarak bu görüntüler üç ESA ağı ile yeniden sınıflandırılmıştır. Bu filtreler uygulandıktan sonra elde edilen sonuçların, orijinal verilerle elde edilen sınıflandırma sonuçlardan daha yüksek olduğu ortaya konmuştur.