Robustness of radiomic features in healthy abdominal parenchyma of patients with repeated examinations on dual-layer dual-energy CT.

OBJECTIVES: Robustness of radiomic features in physiological tissue is an important prerequisite for quantitative analysis of tumor biology and response assessment. In contrast to previous studies which focused on different tumors with mostly short scan-re-scan intervals, this study aimed to evaluate the robustness of radiomic features in cancer-free patients and over a clinically encountered inter-scan interval. MATERIALS AND METHODS: Patients without visible tumor burden who underwent at least two portal-venous phase dual energy CT examinations of the abdomen between May 2016 and January 2020 were included, while macroscopic tumor burden was excluded based upon follow-up imaging for all patients (≥3 months). Further, patients were excluded if no follow-up imaging was available, or if the CT protocol showed deviations between repeated examinations. Circular regions of interest were placed and proofread by two board-certified radiologists (4 years and 5 years experience) within the liver (segments 3 and 6), the psoas muscle (left and right), the pancreatic head, and the spleen to obtain radiomic features from normal-appearing organ parenchyma using PyRadiomics. Radiomic feature robustness was tested using the concordance correlation coefficient with a threshold of 0.75 considered indicative for deeming a feature robust. RESULTS: In total, 160 patients with 480 repeated abdominal CT examinations (range: 2-4 per patient) were retrospectively included in this single-center, IRB-approved study. Considering all organs and feature categories, only 4.58 % (25/546) of all features were robust with the highest rate being found in the first order feature category (20.37 %, 22/108). Other feature categories (grey level co-occurrence matrix, grey level dependence matrix, grey level run length matrix, grey level size zone matrix, and neighborhood gray-tone difference matrix) yielded an overall low percentage of robust features (range: 0.00 %-1.19 %). A subgroup analysis revealed the reconstructed field of view and the X-ray tube current as determinants of feature robustness (significant differences in subgroups for all organs, p < 0.001) as well as the size of the region of interest (no significant difference for the pancreatic head with p = 0.135, significant difference with p < 0.001 for all other organs). CONCLUSION: Radiomic feature robustness obtained from cancer-free subjects with repeated examinations using a consistent protocol and CT scanner was limited, with first order features yielding the highest proportion of robust features.

CNN application for automated determination of the patient's size to obtain the size-specific dose estimated in CT.

Introduction. The currently available dosimetry techniques in computed tomography can be inaccurate which overestimate the absorbed dose. Therefore, we aimed to provide an automated and fast methodology to more accurately calculate the SSDE usingDwobtained by using CNN from thorax and abdominal CT study images.Methods. The SSDE was determined from the 200 records files. For that purpose, patients' size was measured in two ways: (a) by developing an algorithm following the AAPM Report No. 204 methodology; and (b) using a CNN according to AAPM Report No. 220.Results. The patient's size measured by the in-house software in the region of thorax and abdomen was 27.63 ± 3.23 cm and 28.66 ± 3.37 cm, while CNN was 18.90 ± 2.6 cm and 21.77 ± 2.45 cm. The SSDE in thorax according to 204 and 220 reports were 17.26 ± 2.81 mGy and 23.70 ± 2.96 mGy for women and 17.08 ± 2.09 mGy and 23.47 ± 2.34 mGy for men. In abdomen was 18.54 ± 2.25 mGy and 23.40 ± 1.88 mGy in women and 18.37 ± 2.31 mGy and 23.84 ± 2.36 mGy in men.Conclusions. Implementing CNN-based automated methodologies can contribute to fast and accurate dose calculations, thereby improving patient-specific radiation safety in clinical practice.

Protocol optimization for abdominal imaging using photon-counting CT: a consensus of two academic institutions.

PURPOSE: Photon-counting detector CT (PCD CT) is a promising technology for abdominal imaging due to its ability to provide high spatial and contrast resolution images with reduced patient radiation exposure. However, there is currently no consensus regarding the optimal imaging protocols for PCD CT. This article aims to present the PCD CT abdominal imaging protocols used by two tertiary care academic centers in the United States. METHODS: A review of PCD CT abdominal imaging protocols was conducted by two abdominal radiologists at different academic institutions. Protocols were compared in terms of acquisition parameters and reconstruction settings. Both imaging centers independently selected similar protocols for PCD CT abdominal imaging, using QuantumPlus mode. RESULTS: There were some differences in the use of reconstruction kernels and iterative reconstruction levels, however the individual combination at each site resulted in similar image impressions. Overall, the imaging protocols used by both centers provide high-quality images with low radiation exposure. CONCLUSION: These findings provide valuable insights into the development of standardized protocols for PCD CT abdominal imaging, which can help to ensure consistent as well as high-quality imaging across different institutions and allow for future multicenter research collaborations.

Pretreatment Interstitial Lung Abnormalities Detected on Abdominal Computed Tomography Scans in Prostate Cancer Patients.

OBJECTIVE: Prostate cancer and interstitial lung abnormality (ILA) share similar risk factor, which is men and older age. The purpose of this study was to investigate the prevalence of pretreatment ILA among prostate cancer patients who underwent abdominal computed tomography (CT) within 1 year at their first visit to the urology department. In addition, we aimed to assess the association between pretreatment ILA and long-term survival in prostate cancer patients. METHODS: This study was conducted in patients who had a first visit for prostate cancer at urology department between 2005 and 2016 and underwent an abdominal CT within 1 year. A thoracic radiologist evaluated the presence of ILA through inspecting the lung base scanned on an abdominal CT. The association between pretreatment ILA and survival was assessed using Kaplan-Meier analysis with log-rank test. Specific survival rates at 12, 36, and 60 months according to the presence of ILA were evaluated using z -test. Cox regression analysis was used to assess the risk factors of mortality. RESULTS: A total of 173 patients were included (mean age, 70.23 ± 7.98 years). Pretreatment ILA was observed in 10.4% of patients. Patients with ILA were more likely to be older and current smokers. Pretreatment ILA was associated with poor survival ( P < 0.001). Age ≥70 years (hazards ratio [HR], 1.98; 95% confidence interval [CI], 1.24-3.16; P = 0.004), metastatic stage (HR, 2.26; 95% CI, 1.36-3.74; P = 0.002), and ILA (HR, 1.96; 95% CI, 1.06-3.60; P = 0.031) were the independent risk factors of mortality. An ILA (HR, 3.94; 95% CI, 1.78-8.72; P = 0.001) was the only independent risk factor of mortality in localized stage prostate cancer patients. CONCLUSIONS: This study provides important insights into the unexplored effect of pretreatment ILA in prostate cancer patients. Pretreatment ILAs were observed considerably in the lung bases scanned on the abdominal CT scans among prostate cancer patients. Furthermore, pretreatment ILAs were the risk factor of mortality. Therefore, lung bases should be routinely inspected in the abdominal CT scans of prostate cancer patients. This result may help clinicians in establishing personalized management strategy of prostate cancer patients.

Yield of abdominal radiographs in children with suspected intussusception; rate of pneumoperitoneum and other abdominal pathology.

OBJECTIVES: Ultrasound is the criterion standard imaging modality for the diagnosis of intussusception. However, to our knowledge the utility of abdominal radiographs to concurrently screen for pneumoperitoneum or other abdominal pathology that could have a similar presentation has not been studied. Our institutional protocol requires the performance of AP supine and left lateral decubitus views of the abdomen prior to ultrasound evaluation for intussusception, providing an opportunity to examine the yield of abdominal radiographs in this setting. Our primary objective was to determine the rate of pneumoperitoneum on screening abdominal radiographs in children undergoing evaluation for intussusception. Our secondary objective was to determine the rate that other clinically significant pathology is found on these screening abdominal radiographs. METHODS: We performed a retrospective chart review of all patients under 6 years of age who had any imaging ordered in our large urban pediatric emergency department to evaluate for suspected intussusception during the calendar years 2018-2020. RESULTS: 1115 patient encounters met our inclusion criteria. Among 1090 who had screening abdominal radiographs, 82 (8%) had findings concerning for intussusception. Of those not concerning for intussusception, 635 (58%) were read as normal, 263 (24%) showed moderate to large stool burden, 107 (10%) showed generalized bowel distention, and 22 (2%) showed abnormal gastric distention. Individually the remainder of all other findings compromised <1% of encounters and included radiopaque foreign body (8), intraabdominal calcification (4), pneumonia/effusion (3), pneumatosis intestinalis, abdominal mass (2), diaphragmatic hernia (1), rib fracture (1), appendicolith (1), feeding tube malposition (1), and bowel wall thickening (1). In one encounter the patient had a bowel perforation with pneumoperitoneum present secondary to ingestion of multiple magnets. CONCLUSIONS: Our study indicates that radiograph-detected pneumoperitoneum is rare in children with suspected intussusception. Constipation is the most common abnormal finding on screening radiographs. Other findings occur in approximately 15% of total cases, some of which require further workup.

Diagnostic Yield of Abdominal Radiographs in the Pediatric Emergency Department.

BACKGROUND AND OBJECTIVES: Abdominal radiographs (ARs) are commonly used in the pediatric emergency department (PED). Their low diagnostic accuracy leads to overuse, excess radiation exposure, and increased resource usage. This study aims to assess the diagnostic yield of ARs in the evaluation of intraabdominal pathology in the PED. METHODS: Retrospective, cross-sectional study of patients aged 0 to 18 years with an AR who visited the PED between 2017 and 2019. Diagnostic yield was analyzed with sensitivity, specificity, positive predictive value, negative predictive value (NPV), and likelihood ratio. RESULTS: A total of 4288 ARs were identified, with a rate of 6%. The overall abnormal AR rate was 31%. The incidences of an abnormal AR in abdominal pain, vomiting, and constipation were 26%, 37%, and 50%, respectively. There was a 13% rate of clinically significant diagnoses. The AR diagnostic yield showed 44% sensitivity, 70% specificity, 17% positive predictive value, and 90% NPV ( P < 0.05). Unadjusted odds ratio analysis of positive AR and abdominal pain, vomiting, and constipation revealed an odds ratio of 0.68 (95% confidence interval [CI], 0.63-0.75), 1.22 (95% CI, 1.06-1.39), and 1.72 (95% CI, 1.54-1.91), respectively. CONCLUSIONS: There is a low rate of intraabdominal pathologic processes that an AR can identify. A normal AR does not change patient management, nor does it reduce the need for further radiologic imaging. Despite a good NPV, the AR is not a useful diagnostic tool in the PED because of its limited ability to rule in or rule out clinically significant diagnoses.

Plain film of the abdomen remains a low sensitivity test in A&E.

BACKGROUND: Plain film abdomens (PFA) are frequently used in the emergency department to help guide the management of patients presenting with abdominal symptoms. A plain film abdomen contributes minimally to clinical scenarios due to low sensitivity and specificity. Is a PFA useful in the emergency setting or does it serve to further complicate decision making? AIM: We hypothesise that PFAs in the emergency department are over utilised to falsely reassure clinicians and patients alike. METHODS: A search of the National Integrated Medical Imaging System (NIMIS) database in an Irish tertiary referral hospital was conducted. All plain film abdominal radiographs requested by the emergency department between 01/01/2022 and 31/08/2022 were identified. Requests where there was suspicion of foreign body were excluded. A retrospective search of the NIMIS database identified subjects who underwent subsequent imaging. RESULTS: A total of 619 abdominal films were deemed suitable for inclusion. These comprised of 338 male and 282 female subjects. Subjects had an average age of 64 years. Fifty-seven per cent of PFAs detected no abnormality. Forty-two per cent of subjects had subsequent imaging. The plain film findings correlated with further imaging in only 15% of cases. One case of ruptured aortic aneurysm and 11 perforations were detected on computerised tomography, none of these cases were evident on abdominal X-ray. CONCLUSION: Plain film abdomen requests are over utilised in the emergency department. PFAs are not sensitive for detecting acute pathology and should not be used to decide if a patient requires further imaging or a full clinical assessment.

Short-term psychosocial outcomes of adding a non-contrast abdominal computed tomography (CT) scan to the thoracic CT within lung cancer screening.

OBJECTIVES: To evaluate psychological, social, and financial outcomes amongst individuals undergoing a non-contrast abdominal computed tomography (CT) scan to screen for kidney cancer and other abdominal malignancies alongside the thoracic CT within lung cancer screening. SUBJECTS AND METHODS: The Yorkshire Kidney Screening Trial (YKST) is a feasibility study of adding a non-contrast abdominal CT scan to the thoracic CT within lung cancer screening. A total of 500 participants within the YKST, comprising all who had an abnormal CT scan and a random sample of one-third of those with a normal scan between 14/03/2022 and 24/08/2022 were sent a questionnaire at 3 and 6 months. Outcomes included the Psychological Consequences Questionnaire (PCQ), the short-form of the Spielberger State-Trait Anxiety Inventory, and the EuroQoL five Dimensions five Levels scale (EQ-5D-5L). Data were analysed using regression adjusting for participant age, sex, socioeconomic status, education, baseline quality of life (EQ-5D-5L), and ethnicity. RESULTS: A total of 380 (76%) participants returned questionnaires at 3 months and 328 (66%) at 6 months. There was no difference in any outcomes between participants with a normal scan and those with abnormal scans requiring no further action. Individuals requiring initial further investigations or referral had higher scores on the negative PCQ than those with normal scans at 3 months (standardised mean difference 0.28 sd, 95% confidence interval 0.01-0.54; P = 0.044). The difference was greater in those with anxiety or depression at baseline. No differences were seen at 6 months. CONCLUSION: Screening for kidney cancer and other abdominal malignancies using abdominal CT alongside the thoracic CT within lung cancer screening is unlikely to cause significant lasting psychosocial or financial harm to participants with incidental findings.

Computed tomographic findings in a canine ovarian teratoma.

A 2-year-old, intact female, Labrador Retriever was referred for progressive abdominal distension, assessed by emergency clinicians as being extrauterine in origin on AFAST. Abdominal radiographs and ultrasound identified a large, lobulated, partially mineralized, soft tissue, mid-abdominal mass and gravid uterus. Contrast-enhanced CT identified a mixed fat to soft tissue attenuating mass with a complex internal mineralized matrix, heterogeneous contrast enhancement, receiving blood from the left ovarian artery. Histology confirmed a left ovarian teratoma, diffuse endometrial hyperplasia, and fetal implantation. The patient had a good post-operative outcome for 2 years, but was later diagnosed with primary cranial mediastinal neuroendocrine carcinoma.

Development and clinical feasibility of a reduced-dose radiograph in children for feeding tube placement.

BACKGROUND: Temporary feeding tubes are commonly used but may lead to complications if malpositioned. Radiographs are the gold standard for assessing tube position, but clinician concern over radiation risks may curtail their use. OBJECTIVE: We describe development and use of a reduced dose feeding tube radiograph (RDFTR) targeted for evaluation of feeding tube position. MATERIALS AND METHODS: Age-based abdominal radiograph was adapted to use the lowest mAs setting of 0.32 mAs with field of view between carina and iliac crests. The protocol was tested in DIGI-13 line-pair plates and anthropomorphic phantoms. Retrospective review of initial clinical use compared dose area product (DAP) for RDFTR and routine abdomen, chest, or infant chest and abdomen. Review of RDFTR reports assessed tube visibility, malpositioning, and incidental critical findings. RESULTS: Testing through a line-pair phantom showed loss of spatial resolution from 2.2 line pairs to 0.6 line pairs but preserved visibility of feeding tube tip in RDFTR protocol. DAP comparisons across 23,789 exams showed RDFTR reduced median DAP 72-93% compared to abdomen, 55-78% compared to chest, and 76-79% compared to infant chest and abdomen (p<0.001). Review of 3286 reports showed tube was visible in 3256 (99.1%), malpositioned in airway 8 times (0.2%) and in the esophagus 74 times (2.3%). The tip was not visualized in 30 (0.9%). Pneumothorax or pneumoperitoneum was noted seven times (0.2%) but was expected or spurious in five of these cases. CONCLUSION: RDFTR significantly reduces radiation dose in children with temporary feeding tubes while maintaining visibility of tube tip.

Effective dose and image quality for different patient sizes during AP upper abdominal radiography: A phantom study.

INTRODUCTION: Undertaking medical imaging examinations on obese patients can present practical challenges. Choosing optimal imaging protocols can be difficult, especially when promoting the ALARA principle. The aim of this study was to assess the effects of increasing body part thickness on image quality (IQ) and effective dose (ED) during upper abdominal radiography. A secondary aim was to determine the optimum exposure settings for larger sized patients. METHODS: Underweight, standard, overweight and obese abdomen sizes were simulated using an anthropomorphic upper abdomen phantom, without and with additional fat layers (6, 10 and 16 cm). Phantoms were imaged using a variety of tube potentials (70-110 kVp), automatic exposure control (AEC) and a source-to-image distance of 120 cm. IQ was assessed visually using a relative visual grading analysis (VGA) method. Radiation dose was evaluated by calculating the ED using the Monte Carlo PCXMC 2.0 computer program. RESULTS: IQ values showed a statistical reduction (p = 0.006) with increasing phantom size across all examined tube potentials. The highest IQ scores (3.3, 2.8, 2.5 and 2.2, respectively) were obtained at 70/75 kVp for all phantom thicknesses. As tube potential increased the IQ was also shown to decrease. ED showed a statistically significant increase (p < 0.001) with increasing phantom thicknesses. CONCLUSION: Higher EDs were evident when applying lower tube potentials. Using an AEC with high tube potentials (105/110 kVp) can lead to a considerable decrease in ED with acceptable IQ when undertaking upper abdomen radiography on patients with large body part thicknesses. IMPLICATION FOR PRACTICE: Applying higher values of tube potentials for patients who have a thicker abdomen can lead to decreased ED.

[Dynamic magnetic resonance imaging of the pelvic floor: Technical aspects]. / Dynamische Magnetresonanztomographie des Beckenbodens: Technische Aspekte.

BACKGROUND: Dynamic magnetic resonance imaging (MRI) of the pelvic floor plays a key role in imaging complex pelvic floor dysfunction. High-quality examination is crucial for diagnostic benefit but can be technically challenging. OBJECTIVES: The most important technical aspects (patient selection, patient preparation, MRI technology, MRI scan protocol, success control) for obtaining a state-of-the-art dynamic MRI of the pelvic floor are summarized. MATERIALS AND METHODS: Review of the scientific literature on dynamic pelvic MR imaging with special consideration of the joint recommendations provided by the expert panels of European Society of Urogenital Radiology/European Society of Gastrointestinal and Abdominal Radiology (ESUR/ESGAR) in 2016 and Society of Abdominal Radiology (SAR) in 2019. RESULTS: Examination with at least 1.5 T and a surface coil after rectal instillation of ultrasound gel is clinical standard. Dynamic MRI in a closed magnet with the patient in supine position is the most widespread technique. No clinically significant pathologies of the pelvic floor are missed compared to the sitting position in an open magnet. The minimum scan protocol should encompass static, high-resolution T2-imaging (i.e., T2-TSE) in three planes and dynamic sequences with high temporal resolution in sagittal (and possibly axial) plane (i.e., steady-state or balanced steady-state free precession) during squeezing, straining and evacuation. Detailed patient instruction and practicing prior to the scan improve patients' compliance and hence diagnostic quality. CONCLUSIONS: A technically flawless dynamic MRI of the pelvic floor according to these standards can provide information missed by other imaging modalities and hence alter therapeutic strategies.

Unmasking the Hidden Fetus: Introducing the Circled Delta Sign and Reversed Circled Delta Sign for Accurate Identification on Abdominal Radiography.

BACKGROUND Ionizing radiation poses potential health risk to fetuses, including growth retardation, organ malformations, neurological effects, fetal death, and cancer. Therefore, pregnant women rarely undergo X-ray imaging, unless absolutely necessary. However, they can be inadvertently exposed to X-rays while undergoing an examination without being aware of their pregnancy. Given that the likelihood of a fetus appearing on an abdominal radiograph is low, physicians can be unfamiliar with the appearance of a fetus on such images. If a fetus incidentally appears on an abdominal radiograph, the clinician can encounter difficulties in identifying the fetus, potentially leading to unexpected harm during subsequent imaging procedures. CASE REPORT A 41-year-old woman presented with symptoms of postprandial abdominal pain, abdominal fullness, and urinary frequency. Abdominal radiography revealed a large mass with calcifications in the pelvis and lower abdomen, raising suspicion of a large pelvic tumor, such as a teratoma. However, subsequent ultrasound unexpectedly revealed that the woman was actually 6 months pregnant. The presumed tumor was an enlarged uterus, and the presumed calcifications were the fetus's bones. Before X-ray, the woman had denied being pregnant, mistakenly attributing her condition to excess weight and irregular menstrual cycles. Fortunately, the use of ultrasound instead of computed tomography prevented radiation exposure to the fetus, and the baby was delivered in a healthy state at full term. CONCLUSIONS To avoid unintended harm to fetuses, clinicians should consider the possibility of pregnancy when evaluating a pelvic mass in women of childbearing age. To aid clinicians in accurately identifying fetuses on abdominal radiography and thereby reduce the likelihood of misdiagnosis, we propose the "circled delta sign" and the "reversed circled delta sign".

Accurately evaluating for a small bowel obstruction using an abdominal radiograph, by a new method: The Bowel-Spine Ratio.

INTRODUCTION: A well-established method does not exist to rule out a small bowel obstruction using an abdominal xray series with significant accuracy. The hypothesis of the study is that the ratio of an average small bowel diameter to lumbar spine diameter over 0.5 is most likely a small bowel obstruction. METHODS: An x-ray abdominal series measurement technique was applied to 41 subjects with a chief complaint of "abdominal pain" as part of a randomized retrospective case review to predict an obstruction v. non obstruction. A total number of 81 abdominal pain subjects with a mean age of 46.7 years were selected with 40 excluded due to normal small bowel gas pattern. The subject's medical information was unknown to the authors when reading their images. The measurement technique involved averaging the largest and smallest small bowel short axis diameters with comparison to the lowest clearly visible lumbar body width. The subjects' medical course as described in the medical chart or subsequent computed tomography scans were used as the referencing standard to determine presence of obstruction vs non-obstruction. RESULTS: This method, called the Bowel-Spine Ratio (BSR), resulted in a sensitivity of 0.882 (0.622-0.979; 95% CI), specificity of 0.957 (0.760-0.998; 95% CI), accuracy of 94.7% (80.9%-99.1%; 95% CI) and a positive likelihood ratio of 21 for predicting a small bowel obstruction. CONCLUSION: The abdominal series Bowel-Spine Ratio is a simple yet effective technique to screen for a small bowel obstruction using limited resources and to avoid unnecessary computed tomography scans with the potential to reduce health care costs. IMPLICATIONS FOR PRACTICE: Clinicians could have increased confidence in utilizing abdominal radiographs to evaluate for small bowel obstruction.

Unkinking The Kinked: A Rare Intussusception In The Paediatric Emergency Department.

Intussusception is the introversion of the proximal intestinal loop into the distal downstream part of the intestinal loop, leading to gut wall oedema and restriction of blood supply. A high index of suspicion is required to diagnose it timely. Mostly infants aged less than a year are vulnerable to this surgical emergency presenting mostly with abdominal distension, vomiting, reluctance to feed and bright red jelly-like stools which, if left unrecognised, may result in the development of ischaemic portion of the gut and catastrophic outcomes. We present the case of a 60-day-old baby who presented in the Paediatric Emergency Department with progressive abdominal distension and bloody stools. Abdominal X-ray showed dilated bowel and ultrasound imaging showed a target sign positive for intussusception. The baby was rushed to the operating room (OR) due to delayed presentation, where an uneventful exploratory laparotomy was performed. Acute intestinal intussusception remains a cause of low morbidity and mortality rates if recognised earlier.

Imaging Features of Neonatal Bowel Obstruction.

Radiologic evaluation of neonatal bowel obstruction is challenging owing to the overlapping clinical features and imaging appearances of the most common differential diagnoses. The key to providing an appropriate differential diagnosis comes from a combination of the patient's gestational age, clinical features, and imaging findings. While assessment of radiographs can confirm bowel obstruction and indicate whether it is likely proximal or distal, additional findings at upper or lower gastrointestinal contrast study together with use of US are important in providing an appropriate differential diagnosis. The authors provide an in-depth assessment of the appearances of the most common differential diagnoses of proximal and distal neonatal bowel obstruction at abdominal radiography and upper and lower gastrointestinal contrast studies. These are divided into imaging patterns and their associated differential diagnoses on the basis of abdominal radiographic findings. These findings include esophageal atresia variants including the "single bubble," "double bubble," and "triple bubble" and distal bowel obstruction involving the small and large bowel. Entities discussed include esophageal atresia, hypertrophic pyloric stenosis, pyloric atresia, duodenal atresia, duodenal web, malrotation with midgut volvulus, jejunal atresia, ileal atresia, meconium ileus, segmental volvulus, internal hernia, colonic atresia, Hirschsprung disease, and functional immaturity of the large bowel. The authors include the advantages of abdominal US in this algorithm, particularly for hypertrophic pyloric stenosis, duodenal web, malrotation with midgut volvulus, and segmental volvulus. ©RSNA, 2023 Online supplemental material is available for this article. Quiz questions for this article are available through the Online Learning Center.

Deep Learning-Enabled Detection of Pneumoperitoneum in Supine and Erect Abdominal Radiography: Modeling Using Transfer Learning and Semi-Supervised Learning.

OBJECTIVE: Detection of pneumoperitoneum using abdominal radiography, particularly in the supine position, is often challenging. This study aimed to develop and externally validate a deep learning model for the detection of pneumoperitoneum using supine and erect abdominal radiography. MATERIALS AND METHODS: A model that can utilize "pneumoperitoneum" and "non-pneumoperitoneum" classes was developed through knowledge distillation. To train the proposed model with limited training data and weak labels, it was trained using a recently proposed semi-supervised learning method called distillation for self-supervised and self-train learning (DISTL), which leverages the Vision Transformer. The proposed model was first pre-trained with chest radiographs to utilize common knowledge between modalities, fine-tuned, and self-trained on labeled and unlabeled abdominal radiographs. The proposed model was trained using data from supine and erect abdominal radiographs. In total, 191212 chest radiographs (CheXpert data) were used for pre-training, and 5518 labeled and 16671 unlabeled abdominal radiographs were used for fine-tuning and self-supervised learning, respectively. The proposed model was internally validated on 389 abdominal radiographs and externally validated on 475 and 798 abdominal radiographs from the two institutions. We evaluated the performance in diagnosing pneumoperitoneum using the area under the receiver operating characteristic curve (AUC) and compared it with that of radiologists. RESULTS: In the internal validation, the proposed model had an AUC, sensitivity, and specificity of 0.881, 85.4%, and 73.3% and 0.968, 91.1, and 95.0 for supine and erect positions, respectively. In the external validation at the two institutions, the AUCs were 0.835 and 0.852 for the supine position and 0.909 and 0.944 for the erect position. In the reader study, the readers' performances improved with the assistance of the proposed model. CONCLUSION: The proposed model trained with the DISTL method can accurately detect pneumoperitoneum on abdominal radiography in both the supine and erect positions.