Neural architecture search for pneumonia diagnosis from chest X-rays

Deep Learning Approaches for Medical Image Analysis and Diagnosis

In addition to enhancing diagnostic accuracy, deep learning techniques offer the potential to streamline workflows, reduce interpretation time, and ultimately improve patient outcomes. The scalability and adaptability of deep learning algorithms enable their deployment across diverse clinical settings, ranging from radiology departments to point-of-care facilities. Furthermore, ongoing research efforts focus on addressing the challenges of data heterogeneity, model interpretability, and regulatory compliance, paving the way for seamless integration of deep learning solutions into routine clinical practice. As the field continues to evolve, collaborations between clinicians, data scientists, and industry stakeholders will be paramount in harnessing the full potential of deep learning for advancing medical image analysis and diagnosis. Furthermore, the integration of deep learning algorithms with other technologies, including natural language processing and computer vision, may foster multimodal medical data analysis and clinical decision support systems to improve patient care. The future of deep learning in medical image analysis and diagnosis is promising. With each success and advancement, this technology is getting closer to being leveraged for medical purposes. Beyond medical image analysis, patient care pathways like multimodal imaging, imaging genomics, and intelligent operating rooms or intensive care units can benefit from deep learning models.

Enhancing pediatric pneumonia diagnosis through masked autoencoders

Pneumonia, an inflammatory lung condition primarily triggered by bacteria, viruses, or fungi, presents distinctive challenges in pediatric cases due to the unique characteristics of the respiratory system and the potential for rapid deterioration. Timely diagnosis is crucial, particularly in children under 5, who have immature immune systems, making them more susceptible to pneumonia. While chest X-rays are indispensable for diagnosis, challenges arise from subtle radiographic findings, varied clinical presentations, and the subjectivity of interpretations, especially in pediatric cases. Deep learning, particularly transfer learning, has shown promise in improving pneumonia diagnosis by leveraging large labeled datasets. However, the scarcity of labeled data for pediatric chest X-rays presents a hurdle in effective model training. To address this challenge, we explore the potential of self-supervised learning, focusing on the Masked Autoencoder (MAE). By pretraining the MAE model on adult chest X-ray images and fine-tuning the pretrained model on a pediatric pneumonia chest X-ray dataset, we aim to overcome data scarcity issues and enhance diagnostic accuracy for pediatric pneumonia. The proposed approach demonstrated competitive performance an AUC of 0.996 and an accuracy of 95.89% in distinguishing between normal and pneumonia. Additionally, the approach exhibited high AUC values (normal: 0.997, bacterial pneumonia: 0.983, viral pneumonia: 0.956) and an accuracy of 93.86% in classifying normal, bacterial pneumonia, and viral pneumonia. This study also investigated the impact of different masking ratios during pretraining and explored the labeled data efficiency of the MAE model, presenting enhanced diagnostic capabilities for pediatric pneumonia.

GrMoNAS: A granularity-based multi-objective NAS framework for efficient medical diagnosis

Neural Architecture Search (NAS) has been widely applied to automate medical image diagnostics. However, traditional NAS methods require significant computational resources and time for performance evaluation. To address this, we introduce the GrMoNAS framework, designed to balance diagnostic accuracy and efficiency using proxy datasets for granularity transformation and multi-objective optimization algorithms. The approach initiates with a coarse granularity phase, wherein diverse candidate neural architectures undergo evaluation utilizing a reduced proxy dataset. This initial phase facilitates the swift and effective identification of architectures exhibiting promise. Subsequently, in the fine granularity phase, a comprehensive validation and optimization process is undertaken for these identified architectures. Concurrently, employing multi-objective optimization and Pareto frontier sorting aims to enhance both accuracy and computational efficiency simultaneously. Importantly, the GrMoNAS framework is particularly suitable for hospitals with limited computational resources. We evaluated GrMoNAS in a range of medical scenarios, such as COVID-19, Skin cancer, Lung, Colon, and Acute Lymphoblastic Leukemia diseases, comparing it against traditional models like VGG16, VGG19, and recent NAS approaches including GA-CNN, EBNAS, NEXception, and CovNAS. The results show that GrMoNAS achieves comparable or superior diagnostic precision, significantly enhancing diagnostic efficiency. Moreover, GrMoNAS effectively avoids local optima, indicating its significant potential for precision medical diagnosis.

RoMIA: a framework for creating Robust Medical Imaging AI models for chest radiographs

Artificial Intelligence (AI) methods, particularly Deep Neural Networks (DNNs), have shown great promise in a range of medical imaging tasks. However, the susceptibility of DNNs to producing erroneous outputs under the presence of input noise and variations is of great concern and one of the largest challenges to their adoption in medical settings. Towards addressing this challenge, we explore the robustness of DNNs trained for chest radiograph classification under a range of perturbations reflective of clinical settings. We propose RoMIA, a framework for the creation of Robust Medical Imaging AI models. RoMIA adds three key steps to the model training and deployment flow: (i) Noise-added training, wherein a part of the training data is synthetically transformed to represent common noise sources, (ii) Fine-tuning with input mixing, in which the model is refined with inputs formed by mixing data from the original training set with a small number of images from a different source, and (iii) DCT-based denoising, which removes a fraction of high-frequency components of each image before applying the model to classify it. We applied RoMIA to create six different robust models for classifying chest radiographs using the CheXpert dataset. We evaluated the models on the CheXphoto dataset, which consists of naturally and synthetically perturbed images intended to evaluate robustness. Models produced by RoMIA show 3%-5% improvement in robust accuracy, which corresponds to an average reduction of 22.6% in misclassifications. These results suggest that RoMIA can be a useful step towards enabling the adoption of AI models in medical imaging applications.