Isa Yildirim

We propose a prompt-conditioned framework built on MedSigLIP that injects textual priors via Feature-wise Linear Modulation (FiLM) and multi-scale pooling. Text prompts condition patch-token features on clinical intent, enabling data-efficient learning and rapid adaptation. The architecture combines global, local, and texture-aware pooling through separate regression heads fused by a lightweight MLP, trained with pairwise ranking loss. Evaluated on the LDCTIQA2023 (a public LDCT quality assessment challenge) with 1,000 training images, we achieve PLCC = 0.9575, SROCC = 0.9561, and KROCC = 0.8301, surpassing the top-ranked published challenge submissions and demonstrating the effectiveness of our prompt-guided approach.

Block-Matching and 3D Filtering (BM3D) exploits non-local self-similarity priors for denoising but relies on fixed parameters. Deep models such as U-Net are more flexible but often lack interpretability and fail to generalize across noise regimes. In this study, we propose Deep Unfolded BM3D (DU-BM3D), a hybrid framework that unrolls BM3D into a trainable architecture by replacing its fixed collaborative filtering with a learnable U-Net denoiser. This preserves BM3D's non-local structural prior while enabling end-to-end optimization. We evaluate DU-BM3D on low-dose CT (LDCT) denoising and show that it outperforms classic BM3D and standalone U-Net across simulated LDCT at different noise levels, yielding higher PSNR and SSIM, especially in high-noise conditions.

Deep learning-based low-dose computed tomography reconstruction methods already achieve high performance on standard image quality metrics like peak signal-to-noise ratio and structural similarity index measure. Yet, they frequently fail to preserve the critical anatomical details needed for diagnostic tasks. This fundamental limitation hinders their clinical applicability despite their high metric scores. We propose a novel task-adaptive reconstruction framework that addresses this gap by incorporating a frozen pre-trained task network as a regularization term in the reconstruction loss function. Unlike existing joint-training approaches that simultaneously optimize both reconstruction and task networks, and risk diverging from satisfactory reconstructions, our method leverages a pre-trained task model to guide reconstruction training while still maintaining diagnostic quality. We validate our framework on a liver and liver tumor segmentation task. Our task-adaptive models achieve Dice scores up to 0.707, approaching the performance of full-dose scans (0.874), and substantially outperforming joint-training approaches (0.331) and traditional reconstruction methods (0.626). Critically, our framework can be integrated into any existing deep learning-based reconstruction model through simple loss function modification, enabling widespread adoption for task-adaptive optimization in clinical practice. Our codes are available at: https://github.com/itu-biai/task_adaptive_ct

In Computed Tomography (CT) imaging, one of the most serious concerns has always been ionizing radiation. Several approaches have been proposed to reduce the dose level without compromising the image quality. With the emergence of deep learning, thanks to the increasing availability of computational power and huge datasets, data-driven methods have recently received a lot of attention. Deep learning based methods have also been applied in various ways to address the low-dose CT reconstruction problem. However, the success of these methods largely depends on the availability of labeled data. On the other hand, recent studies showed that training can be done successfully without the need for labeled datasets. In this study, a training scheme was defined to use low-dose projections as their own training targets. The self-supervision principle was applied in the projection domain. The parameters of a denoiser neural network were optimized through self-supervised training. It was shown that our method outperformed both traditional and compressed sensing-based iterative methods, and deep learning based unsupervised methods, in the reconstruction of analytic CT phantoms and human CT images in low-dose CT imaging. Our method’s reconstruction quality is also comparable to a well-known supervised method.

The regularization of retinal oxygen tension estimation was previously proposed with an assumption that phosphorescence intensity images are corrupted by additive Gaussian noise. Based on this assumption, a regularized least-squares estimate has been shown to be better than a conventional least-squares estimation. However, this assumption is inconsistent with the acquisition process of phosphorescence intensity images acquired using an intensified charge-coupled device camera. Almost the entire acquisition process is governed by the natural aspects of photons. Therefore, a method based on photon counting statistics is more appropriate. In this study, we propose a regularized oxygen tension estimation method based on photon counting statistics and a phosphorescence lifetime imaging model.

3D cross-sectional images of a breast can be obtained by using limited number of projection data taken from a limited angle with the Digital Breast Tomosynthesis (DBT), which is one of the breast cancer imaging techniques. However, the limited scanning angle and the missing data problem in the DBT cause distortions in reconstructed images. In this study, a regularizer (anisotropic total variation-ATV) that can be used with iterative reconstruction methods has been developed with taking into account DBT system geometry for the missing data problem. At the same time, the minimization of the gradient of the developed ATV, was made with l p norm, which was defined as non-convex depending on the p values. The efficiency of the developed algorithm was evaluated visually and quantitatively using CNR measurements with the help of projection data taken from the real DBT system of the CIRS BR3D phantom. The obtained results confirm that the proposed method provides better contrast resolution to image microcalcifications in a breast tissue.

Low-dose CT imaging requires reconstruction from noisy indirect measurements which can be defined as an ill-posed linear inverse problem. In addition to conventional FBP method in CT imaging, recent compressed sensing based methods exploit handcrafted priors which are mostly simplistic and hard to determine. More recently, deep learning (DL) based methods have become popular in medical imaging field. In CT imaging, DL based methods try to learn a function that maps low-dose images to normal-dose images. Although the results of these methods are promising, their success mostly depends on the availability of high-quality massive datasets. In this study, we proposed a method that does not require any training data or a learning process. Our method exploits such an approach that deep convolutional neural networks (CNNs) generate patterns easier than the noise, therefore randomly initialized generative neural networks can be suitable priors to be used in regularizing the reconstruction. In the experiments, the proposed method is implemented with different loss function variants. Both analytical CT phantoms and real-world CT images are used with different views. Conventional FBP method, a popular iterative method (SART), and TV regularized SART are used in the comparisons. We demonstrated that our method with different loss function variants outperforms the other methods both qualitatively and quantitatively.

BACKGROUND: Iterative image reconstruction in Digital Breast Tomosynthesis (DBT) is a developing modality that produces three-dimensional (3D) reconstructed images of a breast to detect suspicious lesions. Algebraic reconstruction technique (ART), one of the iterative image reconstruction methods, was applied to reconstruct 3D data of breast and is becoming as one alternative method for the conventional image reconstruction techniques such as filtered back projection (FBP) in DBT imaging. OBJECTIVE: A new majorization-minimization (MM) algorithm was presented for TV denoising of signals. In the field of DBT, however, the algorithm has not yet been applied. In this study, we proposed a new method of “ART+TV_3D+MM,” which applies (MM) algorithm to the images reconstructed by ART+TV_3D for different imaging dose levels to investigate a possible reduction of radiation dose. METHODS: Projections of a real breast phantom (CD Pasmam 1054) were acquired with a Siemens MAMMOMAT DBT system. The proposed new method was repeated and tested with 3 different radiation dose levels. The quality of the images reconstructed using the proposed new method were compared with those generated by the commonly used FBP method using both qualitative and quantitative assessments. RESULTS: The new method showed superior results in terms of visual assessment, contrast to noise ratios (CNR), full width at half maximum (FWHM) values and 1D profiles compared with FBP of the Siemens MAMMOMAT. CNR values were evaluated for two different region of interests (ROIs). For instance, CNR values of ROI-2 of FBP and of new method were 1.670 and 1.978 at 100 mAs, respectively. Moreover, while CNR value of ROI-1 of FBP at 100 mAs was 0.955, CNR value of ROI-1 of using new method at 100 mAs was 48.163. FWHM values for FBP and the new method were 2.328 and 1.765 at 56 mAs, 2.032 and 1.661 at 100 mAs, and 2.111 and 1.736 at 199 mAs, respectively. CONCLUSIONS: The results support that using the new method of “ART+TV_3D+MM” could help decrease the radiation dose level, which is one of the most critical limitations of DBT imaging.

The second heart sound (S2) consists of aortic (A2) and pulmonary (P2) components which are produced by closures of aortic and pulmonary valves respectively for a healthy subject. There is a time split between these components namely S2 split which has very important diagnostic potential. There have been several methods proposed to measure the S2 split. Analyzing the signal in time- frequency domain is the most common and important phase of these proposed methods. In this work, we used and compared short time Fourier transform (STFT), continuous wavelet transform (CWT) and Wigner-Ville distribution (WVD) in estimating S2 split by tracking the instantaneous frequency (IF) of the S2 signal. Simulated data set were used in our comparisons. The best performance was achieved when WVD was used in estimating the IF.