Farzad Ahmed

Md Tanvir Rouf Shawon, G. M. Shahariar Shibli, Farzad Ahmed et al.

This paper presents a research study on the use of Convolutional Neural Network (CNN), ResNet50, InceptionV3, EfficientNetB0 and NASNetMobile models to efficiently detect brain tumors in order to reduce the time required for manual review of the report and create an automated system for classifying brain tumors. An automated pipeline is proposed, which encompasses five models: CNN, ResNet50, InceptionV3, EfficientNetB0 and NASNetMobile. The performance of the proposed architecture is evaluated on a balanced dataset and found to yield an accuracy of 99.33% for fine-tuned InceptionV3 model. Furthermore, Explainable AI approaches are incorporated to visualize the model's latent behavior in order to understand its black box behavior. To further optimize the training process, a cost-sensitive neural network approach has been proposed in order to work with imbalanced datasets which has achieved almost 4% more accuracy than the conventional models used in our experiments. The cost-sensitive InceptionV3 (CS-InceptionV3) and CNN (CS-CNN) show a promising accuracy of 92.31% and a recall value of 1.00 respectively on an imbalanced dataset. The proposed models have shown great potential in improving tumor detection accuracy and must be further developed for application in practical solutions. We have provided the datasets and made our implementations publicly available at - https://github.com/shahariar-shibli/Explainable-Cost-Sensitive-Deep-Neural-Networks-for-Brain-Tumor-Detection-from-Brain-MRI-Images

Sajib Kumar Saha Joy, Farzad Ahmed, Al Hasib Mahamud et al.

Pandas is defined as a software library which is used for data analysis in Python programming language. As pandas is a fast, easy and open source data analysis tool, it is rapidly used in different software engineering projects like software development, machine learning, computer vision, natural language processing, robotics, and others. So a huge interests are shown in software developers regarding pandas and a huge number of discussions are now becoming dominant in online developer forums, like Stack Overflow (SO). Such discussions can help to understand the popularity of pandas library and also can help to understand the importance, prevalence, difficulties of pandas topics. The main aim of this research paper is to find the popularity and difficulty of pandas topics. For this regard, SO posts are collected which are related to pandas topic discussions. Topic modeling are done on the textual contents of the posts. We found 26 topics which we further categorized into 5 board categories. We observed that developers discuss variety of pandas topics in SO related to error and excepting handling, visualization, External support, dataframe, and optimization. In addition, a trend chart is generated according to the discussion of topics in a predefined time series. The finding of this paper can provide a path to help the developers, educators and learners. For example, beginner developers can learn most important topics in pandas which are essential for develop any model. Educators can understand the topics which seem hard to learners and can build different tutorials which can make that pandas topic understandable. From this empirical study it is possible to understand the preferences of developers in pandas topic by processing their SO posts

Noushin Afrin, Md Mahamudul Hasan, Mohammed Fazlay Elahi Safin et al.

Automated License Plate Recognition(ALPR) is a system that automatically reads and extracts data from vehicle license plates using image processing and computer vision techniques. The Goal of LPR is to identify and read the license plate number accurately and quickly, even under challenging, conditions such as poor lighting, angled or obscured plates, and different plate fonts and layouts. The proposed method consists of processing the Bengali low-resolution blurred license plates and identifying the plate's characters. The processes include image restoration using GFPGAN, Maximizing contrast, Morphological image processing like dilation, feature extraction and Using Convolutional Neural Networks (CNN), character segmentation and recognition are accomplished. A dataset of 1292 images of Bengali digits and characters was prepared for this project.

Namu Park, Farzad Ahmed, Zhaoyi Sun et al.

Objective: To evaluate large language models (LLMs) against supervised baselines for fine-grained, lesion-level detection of incidentalomas requiring follow-up, addressing the limitations of current document-level classification systems. Methods: We utilized a dataset of 400 annotated radiology reports containing 1,623 verified lesion findings. We compared three supervised transformer-based encoders (BioClinicalModernBERT, ModernBERT, Clinical Longformer) against four generative LLM configurations (Llama 3.1-8B, GPT-4o, GPT-OSS-20b). We introduced a novel inference strategy using lesion-tagged inputs and anatomy-aware prompting to ground model reasoning. Performance was evaluated using class-specific F1-scores. Results: The anatomy-informed GPT-OSS-20b model achieved the highest performance, yielding an incidentaloma-positive macro-F1 of 0.79. This surpassed all supervised baselines (maximum macro-F1: 0.70) and closely matched the inter-annotator agreement of 0.76. Explicit anatomical grounding yielded statistically significant performance gains across GPT-based models (p < 0.05), while a majority-vote ensemble of the top systems further improved the macro-F1 to 0.90. Error analysis revealed that anatomy-aware LLMs demonstrated superior contextual reasoning in distinguishing actionable findings from benign lesions. Conclusion: Generative LLMs, when enhanced with structured lesion tagging and anatomical context, significantly outperform traditional supervised encoders and achieve performance comparable to human experts. This approach offers a reliable, interpretable pathway for automated incidental finding surveillance in radiology workflows.

Farzad Ahmed, Joniel Augustine Jerome, Meliha Yetisgen et al.

Objective: Clinical documentation contains factual, diagnostic, and management errors that can compromise patient safety. Large language models (LLMs) may help detect and correct such errors, but their behavior under different prompting strategies remains unclear. We evaluate zero-shot prompting, static prompting with random exemplars (SPR), and retrieval-augmented dynamic prompting (RDP) for three subtasks of medical error processing: error flag detection, error sentence detection, and error correction. Methods: Using the MEDEC dataset, we evaluated nine instruction-tuned LLMs (GPT, Claude, Gemini, and OpenAI o-series models). We measured performance using accuracy, recall, false-positive rate (FPR), and an aggregate score of ROUGE-1, BLEURT, and BERTScore for error correction. We also analyzed example outputs to identify failure modes and differences between LLM and clinician reasoning. Results: Zero-shot prompting showed low recall in both detection tasks, often missing abbreviation-heavy or atypical errors. SPR improved recall but increased FPR. Across all nine LLMs, RDP reduced FPR by about 15 percent, improved recall by 5 to 10 percent in error sentence detection, and generated more contextually accurate corrections. Conclusion: Across diverse LLMs, RDP outperforms zero-shot and SPR prompting. Using retrieved exemplars improves detection accuracy, reduces false positives, and enhances the reliability of medical error correction.