Zohreh Azimifar

Prof. Zohreh Azimifar


Research Focus & Metrics

Academic Career Length

24Years (2003-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
2 May 2022 in PeerJ Computer Science
Verified
Journal Article
Open access
PDF file
SKShima Kamyab
Shima Kamyab
ZAZohreh Azimifar
Zohreh Azimifar
RSRasool Sabzi
Rasool Sabzi
PFPaul Fieguth
Paul Fieguth
In this paper we investigate a variety of deep learning strategies for solving inverse problems. We classify existing deep learning solutions for inverse problems into three categories of Direct Mapping, Data Consistency Optimizer, and Deep Regularizer. We choose a sample of each inverse problem type, so as to compare the robustness of the three categories, and report a statistical analysis of their differences. We perform extensive experiments on the classic problem of linear regression and three well-known inverse problems in computer vision, namely image denoising, 3D human face inverse rendering, and object tracking, in presence of noise and outliers, are selected as representative prototypes for each class of inverse problems. The overall results and the statistical analyses show that the solution categories have a robustness behaviour dependent on the type of inverse problem domain, and specifically dependent on whether or not the problem includes measurement outliers. Based on our experimental results, we conclude by proposing the most robust solution category for each inverse problem class.

3 April 2025 in BMC Cancer
Verified
Journal Article
Open access
PDF file
ZTZahra Taghados
Zahra Taghados
ZAZohreh Azimifar
Zohreh Azimifar
MMMalihezaman Monsefi
Malihezaman Monsefi
MJMojgan Akbarzadeh Jahromi
Mojgan Akbarzadeh Jahromi
Cervical cancer is a significant global health issue affecting women worldwide, necessitating prompt detection and effective management. According to the World Health Organization (WHO), approximately 660,000 new cases of cervical cancer and 350,000 deaths were reported globally in 2022, with the majority occurring in low- and middle-income countries. These figures emphasize the critical need for effective prevention, early detection, and diagnostic strategies. Recent advancements in machine learning (ML) and deep learning (DL) have greatly enhanced the accuracy of cervical cancer cell classification and diagnosis in manual screening. However, traditional predictive approaches often lack interpretability, which is critical for building explainable AI systems in medicine. Integrating causal reasoning, causal inference, and causal discovery into diagnostic frameworks addresses these challenges by uncovering latent causal relationships rather than relying solely on observational correlations. This ensures greater consistency, comprehensibility, and transparency in medical decision-making. This study introduces CausalCervixNet, a Convolutional Neural Network with Causal Insight (CICNN) tailored for cervical cancer cell classification. By leveraging causality-based methodologies, CausalCervixNet uncovers hidden causal factors in cervical cell images, enhancing both diagnostic accuracy and efficiency. The approach was validated on three datasets: SIPaKMeD, Herlev, and our self-collected ShUCSEIT (Shiraz University-Computer Science, Engineering, and Information Technology) dataset, containing detailed cervical cell cytopathology images. The proposed framework achieved classification accuracies of 99.14%, 97.31%, and 99.09% on the SIPaKMeD, Herlev, and ShUCSEIT datasets, respectively. These results highlight the importance of integrating causal discovery, causal reasoning, and causal inference into diagnostic workflows. By merging causal perspectives with advanced DL models, this research offers an interpretable, reliable, and efficient framework for cervical cancer diagnosis, contributing to improved patient outcomes and advancements in cervical cancer treatment.

18 July 2019 in IEEE Transactions on Intelligent Transportation Systems
Verified
Journal Article
Closed access
ANAmir Nazemi
Amir Nazemi
ZAZohreh Azimifar
Zohreh Azimifar
MSMohammad Javad Shafiee
Mohammad Javad Shafiee
Alexander WongAlexander Wong
Alexander Wong
Vehicle Make and Model Recognition (VMMR) systems provide a fully automatic framework to recognize and classify different vehicle models. Several approaches have been proposed to address this challenge; however, they can perform in restricted conditions. Here, in this paper, we formulate the VMMR as a fine-grained classification problem and propose a new configurable on-road VMMR framework. We benefit from the unsupervised feature learning methods, and in more details, we employ Locality-constraint Linear Coding (LLC) method as a fast feature encoder for encoding the input SIFT features. The proposed method can perform in real environments of different conditions. This framework can recognize 50 models of vehicles and has the advantage to classify every other vehicle not belonging to one of the specified 50 classes as an unknown vehicle. The proposed VMMR framework can be configured to become faster or more accurate based on the application domain. The proposed approach is examined on two datasets, including Iranian on-road vehicle (IORV) dataset and CompuCar dataset. The IORV dataset contains images of 50 models of vehicles captured in real situations by traffic-cameras in different weather and lighting conditions. The experimental results show the advantage of the real-time configuration of the proposed framework over the state-of-the-art methods on the IORV datatset and comparable results on CompuCar dataset with 97.5% and 98.4% accuracies, respectively and acceptable running time.

1 July 2016 in IET Signal Processing
Verified
Journal Article
Open access
PDF file
MYMahdi Yazdian‐Dehkordi
Mahdi Yazdian‐Dehkordi
ZAZohreh Azimifar
Zohreh Azimifar
The GM‐PHD‐based filter has been proposed as an alternative of the PHD filter to estimate the first‐order moment of the multi‐target posterior density. The GM‐PHD filter utilises a weighted summation of Gaussian components to estimate the target states. This filter and its recent variants perform state extraction of the targets based on the target weights. However, due to different uncertainties such as noisy observation, miss‐detection, clutter or occlusion, the weight of a target is decreased and the estimation of the target is lost in some steps. In this study, the authors develop a simple and effective N ‐scan approach which employs the weight history of targets to improve the performance of the GM‐PHD‐based methods. They propose to assign a label, a weight history and a binary confidence indicator to each Gaussian component and propagate them in time. Then, they explain a novel N ‐scan state extraction algorithm to estimate the target states based on their histories in the N last steps. To study the efficiency of the proposed N ‐scan approach, it is applied on the GM‐PHD filter as well as its several recent variants. The experimental results provided for various uncertainties show the effectiveness of the method.

26 March 2021 in Journal of Manufacturing Science and Engineering
Verified
Journal Article
Closed access
PDF file
DEDeniz Sera Ertay
Deniz Sera Ertay
SKShima Kamyab
Shima Kamyab
VVVlasea Vlasea
Vlasea Vlasea
ZAZohreh Azimifar
Zohreh Azimifar
TMThanh Ma
Thanh Ma
PFPaul Fieguth
Paul Fieguth
Achieving defect-free parts is traditionally challenging in laser powder bed fusion (LPBF). The mechanical properties of additively manufactured parts are highly affected by their density; as such, research in defect detection and pore prediction has gained significant interest. The process parameters, the powder characteristics, and the process environment conditions play an important role in defect occurrence. Moreover, the laser scan path affects density, especially at scan path discontinuities. In this work, the complex interaction between the process parameters and the scan path on the occurrence of subsurface pores is investigated. In the data preparation step, a synthetic data set is generated to model the melt pool morphology along the scan path. A secondary data set containing the pore space of the resulting parts is obtained via X-ray computed tomography (CT) and is registered with the synthetic data set. Machine learning models, namely, a Conditional Variational AutoEncoder (CVAE) and a Convolutional Neural Network (CNN), are then trained based on the input features to predict pore occurrence. The performance evaluation of both CNN and CVAE models on synthetic data indicates that the scan path and process parameters can be utilized in predicting pore locations. Quantitative results show that employing offline CT images a priori in training the CVAE, without the need to have CT information in the test phase, leads the CVAE model to superior performance over the CNN.

6 October 2017 in Machine Vision and Applications
Verified
Journal Article
Closed access
PDF file
MFMahmoud Famouri
Mahmoud Famouri
ABAdrien Bartoli
Adrien Bartoli
ZAZohreh Azimifar
Zohreh Azimifar