To evaluate the performance of artificial intelligence (AI)-based coronary artery calcium scoring (CACS) on non-electrocardiogram (ECG)-gated chest CT, using manual quantification as the reference standard, while characterizing per-vessel reliability and clinical risk classification impacts. Retrospective study of 290 patients (June 2023-2024) with paired non-ECG-gated chest CT and ECG-gated cardiac CT (median time was 2 days). AI-based CACS and manual CACS (CACS_man) were compared using intraclass correlation coefficient (ICC) and weighted Cohen's kappa (3,1). Error types, anatomical distributions, and CACS of the lesions of individual arteries or segments were assessed in accordance with the Society of Cardiovascular Computed Tomography (SCCT) guidelines. The total CACS of chest CT demonstrated excellent concordance with CACS_man (ICC = 0.87, 95 % CI 0.84-0.90). Non-ECG-gated chest showed a 7.5-fold increased risk misclassification rate compared to ECG-gated cardiac CT (41.4 % vs. 5.5 %), with 35.5 % overclassification and 5.9 % underclassification. Vessel-specific analysis revealed paradoxical reliability of the left anterior descending artery (LAD) due to stent misclassification in four cases (ICC = 0.93 on chest CT vs 0.82 on cardiac CT), while the right coronary artery (RCA) demonstrated suboptimal performance with ICCs ranging from 0.60 to 0.68. Chest CT exhibited higher false-positive (1.9 % vs 0.5 %) and false-negative rates (14.4 % vs 4.3 %). False positive mainly derived from image noise in proximal LAD/RCA (median CACS 5.97 vs 3.45) and anatomical error, while false negatives involved RCA microcalcifications (median CACS 2.64). AI-based non-ECG-gated chest CT demonstrates utility for opportunistic screening but requires protocol optimization to address vessel-specific limitations and mitigate 41.4 % risk misclassification rates.

Incidental pulmonary emboli (IPE) are frequently overlooked by radiologists. Artificial intelligence (AI) algorithms have been developed to aid detection of pulmonary emboli. To measure diagnostic performance of AI compared with prospective interpretation by radiologists. A commercially available AI algorithm was used to retrospectively review 14,453 contrast-enhanced outpatient CT CAP exams in 9171 patients where PE was not clinically suspected. Natural language processing (NLP) searches of reports identified IPE detected prospectively. Thoracic radiologists reviewed all cases read as positive by AI or NLP to confirm IPE and assess the most proximal level of clot and overall clot burden. 1,400 cases read as negative by both the initial radiologist and AI were re-reviewed to assess for additional IPE. Radiologists prospectively detected 218 IPE and AI detected an additional 36 unreported cases. AI missed 30 cases of IPE detected by the radiologist and had 94 false positives. For 36 IPE missed by the radiologist, median clot burden was 1 and 19 were solitary segmental or subsegmental. For 30 IPE missed by AI, one case had large central emboli and the others were small with 23 solitary subsegmental emboli. Radiologist re-review of 1,400 exams interpreted as negative found 8 additional cases of IPE. Compared with radiologists, AI had similar sensitivity but reduced positive predictive value. Our experience indicates that the AI tool is not ready to be used autonomously without human oversight, but a human observer plus AI is better than either alone for detection of incidental pulmonary emboli.

Intracerebral hemorrhage (ICH) is a medical emergency that demands rapid and accurate diagnosis for optimal patient management. Hemorrhagic lesions' segmentation on CT scans is a necessary first step for acquiring quantitative imaging data that are becoming increasingly useful in the clinical setting. However, traditional manual segmentation is time-consuming and prone to inter-rater variability, creating a need for automated solutions. This study introduces a novel approach combining advanced deep learning models to segment extensive and morphologically variable ICH lesions in non-contrast CT scans. We propose a two-step methodology that begins with a user-defined loose bounding box around the lesion, followed by a fine-tuned YOLOv8-S object detection model to generate precise, slice-specific bounding boxes. These bounding boxes are then used to prompt the Medical Segment Anything Model for accurate lesion segmentation. Our pipeline achieves high segmentation accuracy with minimal supervision, demonstrating strong potential as a practical alternative to task-specific models. We evaluated the model on a dataset of 252 CT scans demonstrating high performance in segmentation accuracy and robustness. Finally, the resulting segmentation tool is integrated into a user-friendly web application prototype, offering clinicians a simple interface for lesion identification and radiomic quantification.

Aneurysmal subarachnoid hemorrhage (SAH) is a life-threatening neurological emergency with mortality rates exceeding 30%. Transfer learning from related hematoma types represents a potentially valuable but underexplored approach. Although Unet architectures remain the gold standard for medical image segmentation due to their effectiveness on limited datasets, Low-Rank Adaptation (LoRA) methods for parameter-efficient transfer learning have been rarely applied to convolutional neural networks in medical imaging contexts. We implemented a Unet architecture pre-trained on computed tomography scans from 124 traumatic brain injury patients across multiple institutions, then fine-tuned on 30 aneurysmal SAH patients from the University of Michigan Health System using 3-fold cross-validation. We developed a novel CP-LoRA method based on tensor CP-decomposition and introduced DoRA variants (DoRA-C, convDoRA, CP-DoRA) that decompose weight matrices into magnitude and directional components. We compared these approaches against existing LoRA methods (LoRA-C, convLoRA) and standard fine-tuning strategies across different modules on a multi-view Unet model. LoRA-based methods consistently outperformed standard Unet fine-tuning. Performance varied by hemorrhage volume, with all methods showing improved accuracy for larger volumes. CP-LoRA achieved comparable performance to existing methods while using significantly fewer parameters. Over-parameterization with higher ranks consistently yielded better performance than strictly low-rank adaptations. This study demonstrates that transfer learning between hematoma types is feasible and that LoRA-based methods significantly outperform conventional Unet fine-tuning for aneurysmal SAH segmentation.

This work presents the results of a methodological transfer from remote sensing to healthcare, adapting AMBER -- a transformer-based model originally designed for multiband images, such as hyperspectral data -- to the task of 3D medical datacube segmentation. In this study, we use the AMBER architecture with Adaptive Fourier Neural Operators (AFNO) in place of the multi-head self-attention mechanism. While existing models rely on various forms of attention to capture global context, AMBER-AFNO achieves this through frequency-domain mixing, enabling a drastic reduction in model complexity. This design reduces the number of trainable parameters by over 80% compared to UNETR++, while maintaining a FLOPs count comparable to other state-of-the-art architectures. Model performance is evaluated on two benchmark 3D medical datasets -- ACDC and Synapse -- using standard metrics such as Dice Similarity Coefficient (DSC) and Hausdorff Distance (HD), demonstrating that AMBER-AFNO achieves competitive or superior accuracy with significant gains in training efficiency, inference speed, and memory usage.

In study, we aimed to dosimetrically evaluate the usability of a new generation autocontouring algorithm (DirectORGANS) that automatically identifies organs and contours them directly in the computed tomography (CT) simulator before creating prostate radiotherapy plans. The CT images of 10 patients were used in this study. The prostates, bladder, rectum, and femoral heads of 10 patients were automatically contoured based on DirectORGANS algorithm at the CT simulator. On the same CT image sets, the same target volumes and contours of organs at risk were manually contoured by an experienced physician using MRI images and used as a reference structure. The doses of manually delineated contours of the target volume and organs at risk and the doses of auto contours of the target volume and organs at risk were obtained from the dose volume histogram of the same plan. Conformity index (CI) and homogeneity index (HI) were calculated to evaluate the target volumes. In critical organ structures, V_60, V_65, V₇₀ for the rectum, V_65, V70, V75, and V₈₀ for the bladder, and maximum doses for femoral heads were evaluated. The Mann-Whitney U test was used for statistical comparison with statistical package SPSS (P < 0.05). Compared to the doses of the manual contours (MC) with auto contours (AC), there was no significant difference between the doses of the organs at risk. However, there were statistically significant differences between HI and CI values due to differences in prostate contouring (P < 0.05). The study showed that the need for clinicians to edit target volumes using MRI before treatment planning. However, it demonstrated that delineating organs at risk was used safely without the need for correction. DirectORGANS algorithm is suitable for use in RT planning to minimize differences between physicians and shorten the duration of this contouring step.

Early detection of lung cancer, which remains one of the leading causes of death worldwide, is important for improved prognosis, and CT scanning is an important diagnostic modality. Lung cancer classification according to CT scan is challenging since the disease is characterized by very variable features. A hybrid deep architecture, ILN-TL-DM, is presented in this paper for precise classification of lung cancer from CT scan images. Initially, an Adaptive Gaussian filtering method is applied during pre-processing to eliminate noise and enhance the quality of the CT image. This is followed by an Improved Attention-based ResU-Net (P-ResU-Net) model being utilized during the segmentation process to accurately isolate the lung and tumor areas from the remaining image. During the process of feature extraction, various features are derived from the segmented images, such as Local Gabor Transitional Pattern (LGTrP), Pyramid of Histograms of Oriented Gradients (PHOG), deep features and improved entropy-based features, all intended to improve the representation of the tumor areas. Finally, classification exploits a hybrid deep learning architecture integrating an improved LeNet structure with Transfer Learning (ILN-TL) and a DeepMaxout (DM) structure. Both model outputs are finally merged with the help of a soft voting strategy, which results in the final classification result that separates cancerous and non-cancerous tissues. The strategy greatly enhances lung cancer detection's accuracy and strength, showcasing how combining sophisticated neural network structures with feature engineering and ensemble methods could be used to achieve better medical image classification. The ILN-TL-DM model consistently outperforms the conventional methods with greater accuracy (0.962), specificity (0.955) and NPV (0.964).

Vertebral compression fractures (VCFs) are the most prevalent skeletal manifestations of osteoporosis in cancer patients. Yet, they are frequently missed or not reported in routine clinical radiology, adversely impacting patient outcomes and quality of life. This study evaluates the diagnostic performance of a deep-learning (DL)-based application and its potential to reduce the miss rate of incidental VCFs in a high-risk cancer population. We retrospectively analysed thoraco-abdomino-pelvic (TAP) CT scans from 1556 patients with stage IV cancer collected consecutively over a 4-month period (September-December 2023) in a tertiary cancer center. A DL-based application flagged cases positive for VCFs, which were subsequently reviewed by two expert radiologists for validation. Additionally, grade 3 fractures identified by the application were independently assessed by two expert interventional radiologists to determine their eligibility for vertebroplasty. Of the 1556 cases, 501 were flagged as positive for VCF by the application, with 436 confirmed as true positives by expert review, yielding a positive predictive value (PPV) of 87%. Common causes of false positives included sclerotic vertebral metastases, scoliosis, and vertebrae misidentification. Notably, 83.5% (364/436) of true positive VCFs were absent from radiology reports, indicating a substantial non-report rate in routine practice. Ten grade 3 fractures were overlooked or not reported by radiologists. Among them, 9 were deemed suitable for vertebroplasty by expert interventional radiologists. This study underscores the potential of DL-based applications to improve the detection of VCFs. The analyzed tool can assist radiologists in detecting more incidental vertebral fractures in adult cancer patients, optimising timely treatment and reducing associated morbidity and economic burden. Moreover, it might enhance patient access to interventional treatments such as vertebroplasty. These findings highlight the transformative role that DL can play in optimising clinical management and outcomes for osteoporosis-related VCFs in cancer patients.

Machine learning models can assist with the selection of electrode arrays required for optimal insertion angles. Cochlea implantation is a successful therapy in patients with severe to profound hearing loss. The effectiveness of a cochlea implant depends on precise insertion and positioning of electrode array within the cochlea, which is known for its variability in shape and size. Preoperative imaging like CT or MRI plays a significant role in evaluating cochlear anatomy and planning the surgical approach to optimize outcomes. In this study, preoperative and postoperative CT and CBCT data of 558 cochlea-implant patients were analyzed in terms of the influence of anatomical factors and insertion depth onto the resulting insertion angle. Machine learning models can predict insertion depths needed for optimal insertion angles, with performance improving by including cochlear dimensions in the models. A simple linear regression using just the insertion depth explained 88% of variability, whereas adding cochlear length or diameter and width further improved predictions up to 94%.