Automatic rib unfolding in postmortem computed tomography: diagnostic evaluation of the OpenRib software compared with the autopsy in the detection of rib fractures

Feasibility of rib fracture detection in low-dose computed tomography images with a large, multicenter datasets-based model

Purpose

To evaluate the feasibility of rib fracture detection in low-dose computed tomography (CT) images with a RetinaNet-based approach and to evaluate the potential of lowdose CT for rib fracture detection compared with regular-dose CT images.

Materials and methods

The RetinaNet-based deep learning model was trained using 7300 scans with 50,410 rib fractures that were used as internal training datasets from four multicenter. The external test datasets consisted of both regular-dose and low-dose chest-abdomen CT images of rib fractures; the MICCAI 2020 RibFrac Challenge Dataset was used as the public dataset. Radiologists' interpretations were used as reference standards. The performance of the model in rib fracture detection was compared with the radiologists' interpretation.

Results

In total, 728 traumatic rib fractures of 100 patients [60 men (60 %); mean age, 53.45 ± 11.19 (standard deviation (SD)); range, 18-77 years] were assessed in the external datasets. In these patients, the regular-dose group had a mean CT dose index volume (CTDIvol) of 7.18 mGy (SD: 2.22) and a mean dose length product (DLP) of 305.38 mGy cm (SD: 95.31); the low-dose group had a mean CTDIvol of 2.79 mGy (SD: 1.11) and a mean DLP of 131.52 mGy cm (SD: 55.58). The sensitivity of the RetinaNet-based model and that of the radiologists was 0.859 and 0.721 in the low-dose CT images and 0.886 and 0.794 in the regular-dose CT images, respectively.

Conclusions

These findings indicate that the RetinaNet-based model can detect rib fractures in low-dose CT images with a robust performance, indicating its feasibility in assisting radiologists with rib fracture diagnosis.

Effects of different virtual monoenergetic CT image data on chest wall post-processing "unfolded ribs" and proposal of an algorithm improvement

PURPOSE

To find out if the use of different virtual monoenergetic data sets enabled by DECT technology might have a negative impact on post-processing applications, specifically in case of the "unfolded ribs" algorithm. Metal or beam hardening artifacts are suspected to generate image artifacts and thus reduce diagnostic accuracy. This paper tries to find out how the generation of "unfolded rib" CT image reformates is influenced by different virtual monoenergetic CT images and looks for possible improvement of the post-processing tool.

MATERIAL AND METHODS

Between March 2021 and April 2021, thin-slice dual-energy CT image data of the chest were used creating "unfolded rib" reformates. The same data sets were analyzed in three steps: first the gold standard with the original algorithm on mixed image data sets followed by the original algorithm on different keV levels (40-120 keV) and finally using a modified algorithm which in the first step used segmentation based on mixed image data sets, followed by segmentation based on different keV levels. Image quality (presence of artifacts), lesion and fracture detectability were assessed for all series.

RESULTS

Both, the original and the modified algorithm resulted in more artifact-free image data sets compared to the gold standard. The modified algorithm resulted in significantly more artifact-free image data sets at the keV-edges (40-120 keV) compared the original algorithm. Especially "black artifacts" and pseudo-lesions, potentially inducing false positive findings, could be reduced in all keV level with the modified algorithm. Detection of focal sclerotic, lytic or mixed (k = 0.990-1.000) lesions was very good for all keV levels. The Fleiss-kappa test for detection of fresh and old rib fractures was ≥ 0.997.

CONCLUSION

The use of different virtual monoenergetic keVs for the "unfolded rib" algorithm is generating different artifacts. Segmentation-based artifacts could be eliminated by the proposed new algorithm, showing the best results at 70-80 keV.

Improving diagnostic performance of rib fractures for the night shift in radiology department using a computer-aided diagnosis system based on deep learning: A clinical retrospective study

OBJECTIVE

To investigate the application value of a computer-aided diagnosis (CAD) system based on deep learning (DL) of rib fractures for night shifts in radiology department.

METHODS

Chest computed tomography (CT) images and structured reports were retrospectively selected from the picture archiving and communication system (PACS) for 2,332 blunt chest trauma patients. In all CT imaging examinations, two on-duty radiologists (radiologists I and II) completed reports using three different reading patterns namely, P1 = independent reading during the day shift; P2 = independent reading during the night shift; and P3 = reading with the aid of a CAD system as the concurrent reader during the night shift. The locations and types of rib fractures were documented for each reading. In this study, the reference standard for rib fractures was established by an expert group. Sensitivity and false positives per scan (FPS) were counted and compared among P1, P2, and P3.

RESULTS

The reference standard verified 6,443 rib fractures in the 2,332 patients. The sensitivity of both radiologists decreased significantly in P2 compared to that in P1 (both p <  0.017). The sensitivities of both radiologists showed no statistical difference between P3 and P1 (both p >  0.017). Radiologist I's FPS increased significantly in P2 compared to P1 (p <  0.017). The FPS of radiologist I showed no statistically significant difference between P3 and P1 (p >  0.017). The FPS of Radiologist II showed no statistical difference among all three reading patterns (p >  0.05).

CONCLUSIONS

DL-based CAD systems can be integrated into the workflow of radiology departments during the night shift to improve the diagnostic performance of CT rib fractures.

Mapping of common rib fracture patterns and the subscapular flail chest associated with operative scapula fractures

BACKGROUND

Rib fractures occur in approximately 10% of trauma patients and are associated with more than 50% of patients with scapula fractures. This study investigates the location and patterns of rib fractures and flail chest occurring in patients with operatively treated scapula fractures. Novel frequency mapping techniques of rib fracture patterns in patients who also injure the closely associated scapula can yield insight into surgical approaches and fixation strategies for complex, multiple injuries patients. We hypothesize that rib fractures have locations of common occurrence when presenting with concomitant scapula fracture that requires operative treatment.

METHODS

Patients with one or more rib fractures and a chest computed tomography scan between 2004 and 2018 were identified from a registry of patients having operatively treated scapula fractures. Unfurled rib images were created using Syngo-CT Bone Reading software (Siemens Inc., Munich, Germany). Rib fracture and flail segment locations were marked and measured for standardized placement on a two-dimensional chest wall template. Location and frequency were then used to create a gradient heat map.

RESULTS

A total of 1,062 fractures on 686 ribs were identified in 86 operatively treated scapula fracture patients. The mean ± SD number of ribs fractured per patient was 8.0 ± 4.1 and included a mean ± SD of 12.3 ± 7.2 total fractures. Rib fractures ipsilateral to the scapula fracture occurred in 96.5% of patients. The most common fracture and flail segment location was ipsilateral and subscapular; 51.4% of rib fractures and 95.7% of flail segments involved ribs 3 to 6.

CONCLUSION

Patients indicated for operative treatment of scapula fractures have a substantial number of rib fractures that tend to most commonly occur posteriorly on the rib cage. There is a pattern of subscapular rib fractures and flail chest adjacent to the thick bony borders of the scapula. This study enables clinicians to better evaluate and diagnose scapular fracture patients with concomitant rib fractures.

LEVEL OF EVIDENCE

Diagnostic test, level IV.

Deep-learning-assisted detection and segmentation of rib fractures from CT scans: Development and validation of FracNet

BACKGROUND

Diagnosis of rib fractures plays an important role in identifying trauma severity. However, quickly and precisely identifying the rib fractures in a large number of CT images with increasing number of patients is a tough task, which is also subject to the qualification of radiologist. We aim at a clinically applicable automatic system for rib fracture detection and segmentation from CT scans.

METHODS

A total of 7,473 annotated traumatic rib fractures from 900 patients in a single center were enrolled into our dataset, named RibFrac Dataset, which were annotated with a human-in-the-loop labeling procedure. We developed a deep learning model, named FracNet, to detect and segment rib fractures. 720, 60 and 120 patients were randomly split as training cohort, tuning cohort and test cohort, respectively. Free-Response ROC (FROC) analysis was used to evaluate the sensitivity and false positives of the detection performance, and Intersection-over-Union (IoU) and Dice Coefficient (Dice) were used to evaluate the segmentation performance of predicted rib fractures. Observer studies, including independent human-only study and human-collaboration study, were used to benchmark the FracNet with human performance and evaluate its clinical applicability. A annotated subset of RibFrac Dataset, including 420 for training, 60 for tuning and 120 for test, as well as our code for model training and evaluation, was open to research community to facilitate both clinical and engineering research.

FINDINGS

Our method achieved a detection sensitivity of 92.9% with 5.27 false positives per scan and a segmentation Dice of 71.5%on the test cohort. Human experts achieved much lower false positives per scan, while underperforming the deep neural networks in terms of detection sensitivities with longer time in diagnosis. With human-computer collobration, human experts achieved higher detection sensitivities than human-only or computer-only diagnosis.

INTERPRETATION

The proposed FracNet provided increasing detection sensitivity of rib fractures with significantly decreased clinical time consumed, which established a clinically applicable method to assist the radiologist in clinical practice.

FUNDING

A full list of funding bodies that contributed to this study can be found in the Acknowledgements section. The funding sources played no role in the study design; collection, analysis, and interpretation of data; writing of the report; or decision to submit the article for publication .

Morphology of direct and indirect rib fractures

Rib fractures are a common finding in legal medicine and information on the impact mechanism is relevant for trauma reconstruction. This study focuses on morphological characteristics of rib fractures resulting from direct or indirect force. Fresh human ribs (n = 312) were divided into two groups and broken through local force (direct) and bending (indirect) in anterolateral areas. The ribs were macerated, visually investigated and the results statistically analysed. The indirect fractures showed a significant larger lateral offset of the internal and external fracture ends while the fracture ends of the direct fractures were more often straight, in line. Also, the morphology of the inner and outer fracture edges was significantly related to fracture type. Direct fractures mostly had rough and jagged inner edges (tension side) and straight, smooth outer edges (compression side), whereas indirect fractures more often showed the characteristics vice versa. The results were more convincing in combination and in ribs from persons aged ≤ 75 years at death. In summary, the direct and indirect rib fractures showed significantly different characteristics regarding orientation and offset of the fracture lines and roughness of the inner and outer fracture edges, which can be helpful to distinguish the traumatizing impact mechanisms in forensic autopsy routine.