RESUMO
Purpose: Determination and development of an effective set of models leveraging Artificial Intelligence techniques to generate a system able to support clinical practitioners working with COVID-19 patients. It involves a pipeline including classification, lung and lesion segmentation, as well as lesion quantification of axial lung CT studies. Approach: A deep neural network architecture based on DenseNet is introduced for the classification of weakly-labeled, variable-sized (and possibly sparse) axial lung CT scans. The models are trained and tested on aggregated, publicly available data sets with over 10 categories. To further assess the models, a data set was collected from multiple medical institutions in Colombia, which includes healthy, COVID-19 and patients with other diseases. It is composed of 1,322 CT studies from a diverse set of CT machines and institutions that make over 550,000 slices. Each CT study was labeled based on a clinical test, and no per-slice annotation took place. This enabled a classification into Normal vs. Abnormal patients, and for those that were considered abnormal, an extra classification step into Abnormal (other diseases) vs. COVID-19. Additionally, the pipeline features a methodology to segment and quantify lesions of COVID-19 patients on the complete CT study, enabling easier localization and progress tracking. Moreover, multiple ablation studies were performed to appropriately assess the elements composing the classification pipeline. Results: The best performing lung CT study classification models achieved 0.83 accuracy, 0.79 sensitivity, 0.87 specificity, 0.82 F1 score and 0.85 precision for the Normal vs. Abnormal task. For the Abnormal vs COVID-19 task, the model obtained 0.86 accuracy, 0.81 sensitivity, 0.91 specificity, 0.84 F1 score and 0.88 precision. The ablation studies showed that using the complete CT study in the pipeline resulted in greater classification performance, restating that relevant COVID-19 patterns cannot be ignored towards the top and bottom of the lung volume. Discussion: The lung CT classification architecture introduced has shown that it can handle weakly-labeled, variable-sized and possibly sparse axial lung studies, reducing the need for expert annotations at a per-slice level. Conclusions: This work presents a working methodology that can guide the development of decision support systems for clinical reasoning in future interventionist or prospective studies.
RESUMO
Background: Recent studies indicate that bladder cancer is among the top 10 most common cancers in the world (Saginala et al. 2022). Bladder cancer frequently reoccurs, and prognostic judgments may vary among clinicians. As a favorable prognosis may help to inform less aggressive treatment plans, classification of histopathology slides is essential for the accurate prognosis and effective treatment of bladder cancer patients. Developing automated and accurate histopathology image analysis methods can help pathologists determine the prognosis of patients with bladder cancer. Materials and methods: In this study, we introduced Bladder4Net, a deep learning pipeline, to classify whole-slide histopathology images of bladder cancer into two classes: low-risk (combination of PUNLMP and low-grade tumors) and high-risk (combination of high-grade and invasive tumors). This pipeline consists of four convolutional neural network (CNN)-based classifiers to address the difficulties of identifying PUNLMP and invasive classes. We evaluated our pipeline on 182 independent whole-slide images from the New Hampshire Bladder Cancer Study (NHBCS) (Karagas et al., 1998; Sverrisson et al., 2014; Sverrisson et al., 2014) collected from 1994 to 2004 and 378 external digitized slides from The Cancer Genome Atlas (TCGA) database (https://www.cancer.gov/tcga). Results: The weighted average F1-score of our approach was 0.91 (95% confidence interval (CI): 0.86-0.94) on the NHBCS dataset and 0.99 (95% CI: 0.97-1.00) on the TCGA dataset. Additionally, we computed Kaplan-Meier survival curves for patients who were predicted as high risk versus those predicted as low risk. For the NHBCS test set, patients predicted as high risk had worse overall survival than those predicted as low risk, with a log-rank p-value of 0.004. Conclusions: If validated through prospective trials, our model could be used in clinical settings to improve patient care.
