Abstract
Chest radiography (CXR) is one of the most used medical imaging modalities. Nevertheless, the interpretation of CXR images is time-consuming and subject to variability. As such, automated systems for pathology detection have been proposed and promising results have been obtained, particularly using deep learning. However, these tools suffer from poor explainability, which represents a major hurdle for their adoption in clinical practice. One proposed explainability method in CXR is through contrastive examples, i.e. by showing an alternative version of the CXR except without the lesion being investigated. While image-level normal/healthy image synthesis has been explored in literature, normal patch synthesis via inpainting has received little attention. In this work, a method to synthesize contrastive examples in CXR based on local synthesis of normal CXR patches is proposed. Based on a contextual attention inpainting network (CAttNet), an anatomically-guided inpainting network (AnaCAttNet) is proposed that leverages anatomical information of the original CXR through segmentation to guide the inpainting for a more realistic reconstruction. A quantitative evaluation of the inpainting is performed, showing that AnaCAttNet outperforms CAttNet (FID of 0.0125 and 0.0132 respectively). Qualitative evaluation by three readers also showed that AnaCAttNet delivers superior reconstruction quality and anatomical realism. In conclusion, the proposed anatomical segmentation module for inpainting is shown to improve inpainting performance.

Abstract
Chest X-ray scans are frequently requested to detect the presence of abnormalities, due to their low-cost and non-invasive nature. The interpretation of these images can be automated to prioritize more urgent exams through deep learning models, but the presence of image artifacts, e.g. lettering, often generates a harmful bias in the classifiers and an increase of false positive results. Consequently, healthcare would benefit from a system that selects the thoracic region of interest prior to deciding whether an image is possibly pathologic. The current work tackles this binary classification exercise, in which an image is either normal or abnormal, using an attention-driven and spatially unsupervised Spatial Transformer Network (STERN), that takes advantage of a novel domain-specific loss to better frame the region of interest. Unlike the state of the art, in which this type of networks is usually employed for image alignment, this work proposes a spatial transformer module that is used specifically for attention, as an alternative to the standard object detection models that typically precede the classifier to crop out the region of interest. In sum, the proposed end-to-end architecture dynamically scales and aligns the input images to maximize the classifier's performance, by selecting the thorax with translation and non-isotropic scaling transformations, and thus eliminating artifacts. Additionally, this paper provides an extensive and objective analysis of the selected regions of interest, by proposing a set of mathematical evaluation metrics. The results indicate that the STERN achieves similar results to using YOLO-cropped images, with reduced computational cost and without the need for localization labels. More specifically, the system is able to distinguish abnormal frontal images from the CheXpert dataset, with a mean AUC of 85.67% -a 2.55% improvement vs. the 0.98% improvement achieved by the YOLO-based counterpart in comparison to a standard baseline classifier. At the same time, the STERN approach requires less than 2/3 of the training parameters, while increasing the inference time per batch in less than 2 ms. Code available via GitHub.

Abstract
Machine Learning models need large amounts of annotated data for training. In the field of medical imaging, labeled data is especially difficult to obtain because the annotations have to be performed by qualified physicians. Natural Language Processing (NLP) tools can be applied to radiology reports to extract labels for medical images automatically. Compared to manual labeling, this approach requires smaller annotation efforts and can therefore facilitate the creation of labeled medical image data sets. In this article, we summarize the literature on this topic spanning from 2013 to 2023, starting with a meta-analysis of the included articles, followed by a qualitative and quantitative systematization of the results. Overall, we found four types of studies on the extraction of labels from radiology reports: those describing systems based on symbolic NLP, statistical NLP, neural NLP, and those describing systems combining or comparing two or more of the latter. Despite the large variety of existing approaches, there is still room for further improvement. This work can contribute to the development of new techniques or the improvement of existing ones.

Abstract
Objective: This study aims to forecast the progression of lung cancer by estimating the future diameter of lung nodules. Methods: This approach uses as input the tabular data, axial images from tomography scans, and both data types, employing a ResNet50 model for image feature extraction and direct analysis of patient information for tabular data. The data are processed through a neural network before prediction. In the training phase, class weights are assigned based on the rarity of different types of nodules within the dataset, in alignment with nodule management guidelines. Results: Tabular data alone yielded the most accurate results, with a mean absolute deviation of 0.99 mm. For malignant nodules, the best performance, marked by a deviation of 2.82 mm, was achieved using tabular data applying Lung-RADS class weights during training. The tabular data results highlight the influence of using the initial nodule size as an input feature. These results surpass the literature reference of 348-day volume doubling time for malignant nodules. Conclusion: The developed predictive model is optimized for integration into a clinical workflow after detecting, segmenting, and classifying nodules. It provides accurate growth forecasts, establishing a more objective basis for determining follow-up intervals. Significance: With lung cancer's low survival rates, the capacity for precise nodule growth prediction represents a significant breakthrough. This methodology promises to revolutionize patient care and management, enhancing the chances for early risk assessment and effective intervention.

Abstract
Lung cancer represents a significant health concern necessitating diligent monitoring of individuals at risk. While the detection of pulmonary nodules warrants clinical attention, not all cases require immediate surgical intervention, often calling for a strategic approach to follow-up decisions. The Lung-RADS guideline serves as a cornerstone in clinical practice, furnishing structured recommendations based on various nodule characteristics, including size, calcification, and texture, outlined within established reference tables. However, the reliance on labor-intensive manual measurements underscores the potential advantages of integrating decision support systems into this process. Herein, we propose a feature-based methodology aimed at enhancing clinical decision-making by automating the assessment of nodules in computed tomography scans. Leveraging algorithms tailored for nodule calcification, texture analysis, and segmentation, our approach facilitates the automated classification of follow-up recommendations aligned with Lung-RADS criteria. Comparison with a previously reported end-to-end image-based classification method revealed competitive performance, with the feature-based approach achieving an accuracy of 0.701 +/- 0.026, while the end-to-end method attained 0.727 +/- 0.020. The inherent explainability of the feature-based approach offers distinct advantages, allowing clinicians to scrutinize and modify individual features to address disagreements or rectify inaccuracies, thereby tailoring follow-up recommendations to patient profiles.