Dataset Information

Automatic intracranial abnormality detection and localization in head CT scans by learning from free-text reports.

ABSTRACT: Deep learning has yielded promising results for medical image diagnosis but relies heavily on manual image annotations, which are expensive to acquire. We present Cross-DL, a cross-modality learning framework for intracranial abnormality detection and localization in head computed tomography (CT) scans by learning from free-text imaging reports. Cross-DL has a discretizer that automatically extracts discrete labels of abnormality types and locations from reports, which are utilized to train an image analyzer by a dynamic multi-instance learning approach. Benefiting from the low annotation cost and a consequent large-scale training set of 28,472 CT scans, Cross-DL achieves accurate performance, with an average area under the receiver operating characteristic curve (AUROC) of 0.956 (95% confidence interval: 0.952-0.959) in detecting 4 abnormality types in 17 regions while accurately localizing abnormalities at the voxel level. An intracranial hemorrhage classification experiment on the external dataset CQ500 achieves an AUROC of 0.928 (0.905-0.951). The model can also help review prioritization.

SUBMITTER: Liu A

PROVIDER: S-EPMC10518589 | biostudies-literature | 2023 Sep

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Automatic intracranial abnormality detection and localization in head CT scans by learning from free-text reports.

Liu Aohan A Guo Yuchen Y Lyu Jinhao J Xie Jing J Xu Feng F Lou Xin X Yong Jun-Hai JH Dai Qionghai Q

Cell reports. Medicine 20230821 9

Deep learning has yielded promising results for medical image diagnosis but relies heavily on manual image annotations, which are expensive to acquire. We present Cross-DL, a cross-modality learning framework for intracranial abnormality detection and localization in head computed tomography (CT) scans by learning from free-text imaging reports. Cross-DL has a discretizer that automatically extracts discrete labels of abnormality types and locations from reports, which are utilized to train an i ...[more]

PMID: 37652014

Similar Datasets

Project description:Objective The aim of this study was to assess efficacy of large language models (LLMs) for converting free-text computed tomography (CT) scan reports of head and neck cancer (HNCa) patients into a structured format using a predefined template. Materials and Methods A retrospective study was conducted using 150 CT reports of HNCa patients. A comprehensive structured reporting template for HNCa CT scans was developed, and the Generative Pre-trained Transformer 4 (GPT-4) was initially used to convert 50 CT reports into a structured format using this template. The generated structured reports were then evaluated by a radiologist for instances of missing or misinterpreted information and any erroneous additional details added by GPT-4. Following this assessment, the template was refined for improved accuracy. This revised template was then used for conversion of 100 other HNCa CT reports into structured format using GPT-4. These reports were then reevaluated in the same manner. Results Initially, GPT-4 successfully converted all 50 free-text reports into structured reports. However, there were 10 places with missing information: tracheostomy tube ( n = 3), noninclusion of involvement of sternocleidomastoid muscle ( n = 2), extranodal tumor extension ( n = 3), and contiguous involvement of the neck structures by nodal mass rather than the primary ( n = 2). Few instances of nonsuspicious lung nodules were misinterpreted as metastases ( n = 2). GPT-4 did not indicate any erroneous additional findings. Using the revised reporting template, GPT-4 converted all the 100 CT reports into a structured format with no repeated or additional mistakes. Conclusion LLMs can be used for structuring free-text radiology reports using plain language prompts and a simple yet comprehensive reporting template. Key Points Structured radiology reports in oncological patients, although advantageous, are not used widely in practice due to perceived drawbacks like interference with routine radiology workflow and scan interpretation.We found that GPT-4 is highly efficient in converting conventional CT reports of HNCa patients to structured reports using a predefined template.This application of LLMs in radiology can help in enhancing the acceptability and clinical utility of structured radiology reports in oncological imaging. Summary Statement Large language models can successfully and accurately convert conventional radiology reports for oncology scans into a structured format using a comprehensive predefined template and thus can enhance the utility and integration of these reports in routine clinical practice.

Project description:ObjectiveWe implement 2 different multitask learning (MTL) techniques, hard parameter sharing and cross-stitch, to train a word-level convolutional neural network (CNN) specifically designed for automatic extraction of cancer data from unstructured text in pathology reports. We show the importance of learning related information extraction (IE) tasks leveraging shared representations across the tasks to achieve state-of-the-art performance in classification accuracy and computational efficiency.Materials and methodsMultitask CNN (MTCNN) attempts to tackle document information extraction by learning to extract multiple key cancer characteristics simultaneously. We trained our MTCNN to perform 5 information extraction tasks: (1) primary cancer site (65 classes), (2) laterality (4 classes), (3) behavior (3 classes), (4) histological type (63 classes), and (5) histological grade (5 classes). We evaluated the performance on a corpus of 95 231 pathology documents (71 223 unique tumors) obtained from the Louisiana Tumor Registry. We compared the performance of the MTCNN models against single-task CNN models and 2 traditional machine learning approaches, namely support vector machine (SVM) and random forest classifier (RFC).ResultsMTCNNs offered superior performance across all 5 tasks in terms of classification accuracy as compared with the other machine learning models. Based on retrospective evaluation, the hard parameter sharing and cross-stitch MTCNN models correctly classified 59.04% and 57.93% of the pathology reports respectively across all 5 tasks. The baseline models achieved 53.68% (CNN), 46.37% (RFC), and 36.75% (SVM). Based on prospective evaluation, the percentages of correctly classified cases across the 5 tasks were 60.11% (hard parameter sharing), 58.13% (cross-stitch), 51.30% (single-task CNN), 42.07% (RFC), and 35.16% (SVM). Moreover, hard parameter sharing MTCNNs outperformed the other models in computational efficiency by using about the same number of trainable parameters as a single-task CNN.ConclusionsThe hard parameter sharing MTCNN offers superior classification accuracy for automated coding support of pathology documents across a wide range of cancers and multiple information extraction tasks while maintaining similar training and inference time as those of a single task-specific model.

Project description:BackgroundArtificial intelligence (AI) systems designed to detect abnormalities in abdominal computed tomography (CT) could reduce radiologists' workload and improve diagnostic processes. However, development of such models has been hampered by the shortage of large expert-annotated datasets. Here, we used information from free-text radiology reports, rather than manual annotations, to develop a deep-learning-based pipeline for comprehensive detection of abdominal CT abnormalities.MethodsIn this multicentre retrospective study, we developed a deep-learning-based pipeline to detect abnormalities in the liver, gallbladder, pancreas, spleen, and kidneys. Abdominal CT exams and related free-text reports obtained during routine clinical practice collected from three institutions were used for training and internal testing, while data collected from six institutions were used for external testing. A multi-organ segmentation model and an information extraction schema were used to extract specific organ images and disease information, CT images and radiology reports, respectively, which were used to train a multiple-instance learning model for anomaly detection. Its performance was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, and F1 score against radiologists' ground-truth labels.FindingsWe trained the model for each organ on images selected from 66,684 exams (39,255 patients) and tested it on 300 (295 patients) and 600 (596 patients) exams for internal and external validation, respectively. In the external test cohort, the overall AUC for detecting organ abnormalities was 0.886. Whereas models trained on human-annotated labels performed better with the same number of exams, those trained on larger datasets with labels auto-extracted via the information extraction schema significantly outperformed human-annotated label-derived models.InterpretationUsing disease information from routine clinical free-text radiology reports allows development of accurate anomaly detection models without requiring manual annotations. This approach is applicable to various anatomical sites and could streamline diagnostic processes.FundingJapan Science and Technology Agency.

Project description:BackgroundBolus tracking can optimize the time delay between contrast injection and diagnostic scan initiation in contrast-enhanced computed tomography (CT), yet the procedure is time-consuming and subject to inter- and intra-operator variances which affect the enhancement levels in diagnostic scans. The objective of the current study is to use artificial intelligence algorithms to fully automate the bolus tracking procedure in contrast-enhanced abdominal CT exams for improved standardization and diagnostic accuracy while providing a simplified imaging workflow.MethodsThis retrospective study used abdominal CT exams collected under a dedicated Institutional Review Board (IRB). Input data consisted of CT topograms and images with high heterogeneity in terms of anatomy, sex, cancer pathologies, and imaging artifacts acquired with four different CT scanner models. Our method consisted of two sequential steps: (I) automatic locator scan positioning on topograms, and (II) automatic region-of-interest (ROI) positioning within the aorta on locator scans. The task of locator scan positioning is formulated as a regression problem, where the limited amount of annotated data is circumvented using transfer learning. The task of ROI positioning is formulated as a segmentation problem.ResultsOur locator scan positioning network offered improved positional consistency compared to a high degree of variance in manual slice positionings, verifying inter-operator variance as a significant source of error. When trained using expert-user ground-truth labels, the locator scan positioning network achieved a sub-centimeter error (9.76±6.78 mm) on a test dataset. The ROI segmentation network achieved a sub-millimeter absolute error (0.99±0.66 mm) on a test dataset.ConclusionsLocator scan positioning networks offer improved positional consistency compared to manual slice positionings and verified inter-operator variance as an important source of error. By significantly reducing operator-related decisions, this method opens opportunities to standardize and simplify the workflow of bolus tracking procedures for contrast-enhanced CT.

Dataset Information

Automatic intracranial abnormality detection and localization in head CT scans by learning from free-text reports.

Publications

Automatic intracranial abnormality detection and localization in head CT scans by learning from free-text reports.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets