Performance of deep learning-based algorithm for detection of ileocolic intussusception on abdominal radiographs of young children

Kim, Sung-Won; Yoon, Haesung; Lee, Mi Jung; Kim, Myung Joon; Han, Kyunghwa; Yoon, Ja Kyung; Kim, Hyung Cheol; Shin, Jaeseung; Shin, Hyun Joo

doi:10.1038/s41598-019-55536-6

Cited by 14 publications

(18 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Kim et al reported that drawing rectangular ROI indicators on abdominal radiographs could allow deep learning-based algorithms to aid in screening for right upper quadrant ileocolic intussusception in young patients. According to a 75-image internal validation test, the sensitivity and specificity values of their algorithm were 0.76 and 0.96, respectively, which are better than those of a radiologist who was found to have sensitivity and specificity values of 0.56 and 0.92, respectively 23 . In our study, we drew a rectangular ROI that encompassed the entire abdomen; the ranges of the sensitivity and specificity values after conducting training and internal tests using two data sets were 0.913-0.943 and 0.851-0.905, respectively.…”

Section: Discussionmentioning

confidence: 90%

Deep learning algorithms for detecting and visualising intussusception on plain abdominal radiography in children: a retrospective multicenter study

Kwon

Ryu

et al. 2020

Sci Rep

View full text Add to dashboard Cite

This study aimed to verify a deep convolutional neural network (CNN) algorithm to detect intussusception in children using a human-annotated data set of plain abdominal X-rays from affected children. From January 2005 to August 2019, 1449 images were collected from plain abdominal X-rays of patients ≤ 6 years old who were diagnosed with intussusception while 9935 images were collected from patients without intussusception from three tertiary academic hospitals (A, B, and C data sets). Single Shot MultiBox Detector and ResNet were used for abdominal detection and intussusception classification, respectively. The diagnostic performance of the algorithm was analysed using internal and external validation tests. The internal test values after training with two hospital data sets were 0.946 to 0.971 for the area under the receiver operating characteristic curve (AUC), 0.927 to 0.952 for the highest accuracy, and 0.764 to 0.848 for the highest Youden index. The values from external test using the remaining data set were all lower (P-value < 0.001). The mean values of the internal test with all data sets were 0.935 and 0.743 for the AUC and Youden Index, respectively. Detection of intussusception by deep CNN and plain abdominal X-rays could aid in screening for intussusception in children.

show abstract

Section: Discussionmentioning

confidence: 90%

Deep learning algorithms for detecting and visualising intussusception on plain abdominal radiography in children: a retrospective multicenter study

Kwon

Ryu

et al. 2020

Sci Rep

View full text Add to dashboard Cite

show abstract

“…As mentioned in related work, there are several other technical options to achieve a similar goal of automatic diagnosis of certain disease from radiograph images, including feature point, active shape model, and full machine learning, which not only recognizes region of interest but also performs the measurements automatically [ 11 , 12 , 18 ]. We used a two-phase algorithm, which separated machine-learning-based region of interest recognition and parameter measurement because doing so has several advantages compared with the other options.…”

Section: Discussionmentioning

confidence: 99%

“…The femoral axis and tibial axis will also be determined. Then, YOLOv3 ( https://pjreddie.com/darknet/yolo/ ), an object-detection tool based on machine learning, is used to segment the knee joint from the original image [ 11 , 12 ]. A median filter, i.e., block-matching and three-dimensional (BM3D) filter, and histogram equalization will be applied to the segmented image for enhancement and removal of noise.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Automatic Discoid Lateral Meniscus Diagnosis from Radiographs Based on Image Processing Tools and Machine Learning

Sun

Jiao

et al. 2021

Journal of Healthcare Engineering

View full text Add to dashboard Cite

The aim of the present study is to build a software implementation of a previous study and to diagnose discoid lateral menisci on knee joint radiograph images. A total of 160 images from normal individuals and patients who were diagnosed with discoid lateral menisci were included. Our software implementation includes two parts: preprocessing and measurement. In the first phase, the whole radiograph image was analyzed to obtain basic information about the patient. Machine learning was used to segment the knee joint from the original radiograph image. Image enhancement and denoising tools were used to strengthen the image and remove noise. In the second phase, edge detection was used to quantify important features in the image. A specific algorithm was designed to build a model of the knee joint and measure the parameters. Of the test images, 99.65% were segmented correctly. Furthermore, 97.5% of the tested images were segmented correctly and their parameters were measured successfully. There was no significant difference between manual and automatic measurements in the discoid ( P = 0.28 ) and control groups ( P = 0.15 ). The mean and standard deviations of the ratio of lateral joint space distance to the height of the lateral tibial spine were compared with the results of manual measurement. The software performed well on raw radiographs, showing a satisfying success rate and robustness. Thus, it is possible to diagnose discoid lateral menisci on radiographs with the help of radiograph-image-analyzing software (BM3D, etc.) and artificial intelligence-related tools (YOLOv3). The results of this study can help build a joint database that contains data from patients and thus can play a role in the diagnosis of discoid lateral menisci and other knee joint diseases in the future.

show abstract

“…Recently, Kim et al [ 22 ] have evaluated diagnostic performances of artificial intelligence (AI) and a deep learning-based algorithm for detecting ileocolic intussusception using AXR. Our study was meaningful as a pilot study for developing AI deep learning algorithm, since it determined AXR's associations with recurrent intussusceptions after therapeutic reduction.…”

Section: Discussionmentioning

confidence: 99%

Suspecting Intussusception and Recurrence Risk Stratification Using Clinical Data and Plain Abdominal Radiographs

et al. 2021

Pediatr Gastroenterol Hepatol Nutr

View full text Add to dashboard Cite

Purpose Although ultrasonography is the gold standard of diagnosing intussusception, plain abdomen radiograph (AXR) is often used to make differential diagnosis for pediatric patients with abdominal pain. In intussusception patients, we aimed to analyze the AXR and clinical data to determine the characteristics of early AXR findings associated with diagnosis of intussusception and recurrence after reduction. Methods Between January 2011 and June 2018, 446 patients diagnosed with intussusception based on International Classification of Diseases-10 code of K56.1 were admitted. We retrospectively reviewed medical records of 398 patients who received air reduction; 51 of them have recurred after initial reduction. We evaluated six AXR features including absent ascending colon gas, absent transverse colon gas, target sign, meniscus sign, mass, and ileus. Clinical data and AXR features were compared between single episode and recurrence groups. Results Two groups did not show significant differences regarding clinical data. Mean time to recurrence from air reduction was 3.4±3.2 days. Absent ascending colon gas (63.9%) was the most common feature in intussusception, followed by mass (29.1%). All of six AXR features were observed more frequently in the recurrence group. Absent transverse colon gas was the most closely associated AXR finding for recurrence (odds ratio, 2.964; 95% confidence interval, 1.327–6.618; p =0.008). Conclusion In our study, absence of ascending colon gas was the most frequently seen AXR factor in intussusception patients. Extended and careful observation after reduction may be beneficial if such finding on AXR is found in intussusception patients.

show abstract

Performance of deep learning-based algorithm for detection of ileocolic intussusception on abdominal radiographs of young children

Cited by 14 publications

References 26 publications

Deep learning algorithms for detecting and visualising intussusception on plain abdominal radiography in children: a retrospective multicenter study

Deep learning algorithms for detecting and visualising intussusception on plain abdominal radiography in children: a retrospective multicenter study

Automatic Discoid Lateral Meniscus Diagnosis from Radiographs Based on Image Processing Tools and Machine Learning

Suspecting Intussusception and Recurrence Risk Stratification Using Clinical Data and Plain Abdominal Radiographs

Contact Info

Product

Resources

About