Automatic Detection of Pancreatic Lesions and Main Pancreatic Duct Dilatation on Portal Venous CT Scans Using Deep Learning

OBJECTIVES

This study proposes and evaluates a deep learning method to detect pancreatic neoplasms and to identify main pancreatic duct (MPD) dilatation on portal venous computed tomography scans.

MATERIALS AND METHODS

A total of 2890 portal venous computed tomography scans from 9 institutions were acquired, among which 2185 had a pancreatic neoplasm and 705 were healthy controls. Each scan was reviewed by one in a group of 9 radiologists. Physicians contoured the pancreas, pancreatic lesions if present, and the MPD if visible. They also assessed tumor type and MPD dilatation. Data were split into a training and independent testing set of 2134 and 756 cases, respectively.A method to detect pancreatic lesions and MPD dilatation was built in 3 steps. First, a segmentation network was trained in a 5-fold cross-validation manner. Second, outputs of this network were postprocessed to extract imaging features: a normalized lesion risk, the predicted lesion diameter, and the MPD diameter in the head, body, and tail of the pancreas. Third, 2 logistic regression models were calibrated to predict lesion presence and MPD dilatation, respectively. Performance was assessed on the independent test cohort using receiver operating characteristic analysis. The method was also evaluated on subgroups defined based on lesion types and characteristics.

RESULTS

The area under the curve of the model detecting lesion presence in a patient was 0.98 (95% confidence interval [CI], 0.97-0.99). A sensitivity of 0.94 (469 of 493; 95% CI, 0.92-0.97) was reported. Similar values were obtained in patients with small (less than 2 cm) and isodense lesions with a sensitivity of 0.94 (115 of 123; 95% CI, 0.87-0.98) and 0.95 (53 of 56, 95% CI, 0.87-1.0), respectively. The model sensitivity was also comparable across lesion types with values of 0.94 (95% CI, 0.91-0.97), 1.0 (95% CI, 0.98-1.0), 0.96 (95% CI, 0.97-1.0) for pancreatic ductal adenocarcinoma, neuroendocrine tumor, and intraductal papillary neoplasm, respectively. Regarding MPD dilatation detection, the model had an area under the curve of 0.97 (95% CI, 0.96-0.98).

CONCLUSIONS

The proposed approach showed high quantitative performance to identify patients with pancreatic neoplasms and to detect MPD dilatation on an independent test cohort. Performance was robust across subgroups of patients with different lesion characteristics and types. Results confirmed the interest to combine a direct lesion detection approach with secondary features such as the MPD diameter, thus indicating a promising avenue for the detection of pancreatic cancer at early stages.

EGS P08 Endoscopic ultrasound-guided transmural gallbladder drainage – Who can benefit?

Background

Endoscopic ultrasound-guided gallbladder drainage (EGBD, cholecysto-duodenostomy) is a safe alternative to percutaneous gallbladder drainage (PGBD) for treatment of acute cholecystitis, and avoids morbidity from external drains. However it may complicate attempts at future cholecystectomy. We aimed to identify a patient subgroup that would not be fit for surgery in the future and therefore should be considered for EGBD.

Methods

Patients with cholecystitis were identified retrospectively using hospital coding (2012–2022). Additional codes for cholecystectomy and percutaneous drainage were used to identify subgroups. Patient characteristics were also collected. Charleston comorbidity index (CCI) and APACHEII score were calculated where data were available. Statistical analyses were performed in R (Welch t-test). Data are displayed (mean LC vs mean NLC, p value).

Results

5653 patients with cholecystitis were identified (3765 LC, 66.6%). 53(0.94%) underwent PGBD. 23(43%) subsequently underwent LC. Patients in the LC group were younger (62.2vs74 p=0.002) and had a lower CCI (3.6vs5.9 p=0.001) compared to the non-operative (NLC) group. APACHE scores were calculated in 40/53 patients (16 LC, 24 NLC) and were similar between groups (8.5vs10.6 p=0.76). BMI was calculated in 41 patients (18 LC, 23 NLC) and was not significantly different (30.0vs30.6 p=0.19).

Conclusions

We have identified an older, more comorbid subgroup of patients with severe cholecystitis who are unlikely to be fit for surgery and could potentially benefit from definitive EGBD. There remains a place for PGBD in patients that will eventually become suitable for LC.

Electrocardiography-gated CT for acute aortic syndrome: quantifying the potential impact of subspecialty national recommendations on emergency general radiology reporting

AIM

To evaluate the detection of acute aortic syndrome (AAS) and the prevalence of alternative diagnoses that may explain the presentation or require follow-up.

MATERIALS AND METHODS

This was a retrospective, blinded re-evaluation of consecutive electrocardiography (ECG)-gated computed tomography (CT) aortic studies by a cardiovascular radiologist performed between September 2019 and May 2020 in a tertiary-referral cardiothoracic centre.

RESULTS

There were 118 identified examinations, six examinations were excluded leaving 112 (mean age = 61 ± 17; 56% male). Three cases of AAS were present (prevalence 2.7%); only one was reported on initial review. There were no false-positive diagnoses of AAS. The heart was mentioned in 79 (70.5%) reports and 73 (65.2%) of reviews revealed a total of 114 new observations; 111 (97.4%) of these were cardiovascular with 44/112 (39.3%) patients potentially having a significant previously unsuspected cardiovascular diagnosis.

CONCLUSION

The implementation of national clinical guidance to increase testing and improve image quality led to a series of challenges. The real value of ECG-gated CT may lie in detecting other diseases that mimic AAS. With the additional workload, increased subspecialty expertise is required but there needs to be a willingness to learn with an adequate support infrastructure.

Spontaneous portomesenteric thrombosis in a non-cirrhotic patient with SARS-CoV-2 infection

Intra-abdominal thromboses are a poorly characterised thrombotic complication of COVID-19 and are illustrated in this case. A 42-year-old man with chronic hepatitis B (undetectable viral load, FibroScan 7.4 kPa) developed fever and cough in March 2020. 14 days later, he developed right upper quadrant pain. After being discharged with reassurance, he re-presented with worsening pain on symptom day 25. Subsequent abdominal ultrasound suggested portal vein thrombosis. CT of the abdomen confirmed portal and mid-superior mesenteric vein thromboses. Concurrent CT of the chest suggested COVID-19 infection. While reverse transcription PCR was negative, subsequent antibody serology was positive. Thrombophilia screen excluded inherited and acquired thrombophilia. Having been commenced on apixaban 5 mg two times per day, he is currently asymptomatic. This is the first case of COVID-19-related portomesenteric thrombosis described in the UK. A recent meta-analysis suggests 9.2% of COVID-19 cases develop abdominal pain. Threshold for performing abdominal imaging must be lower to avoid this reversible complication.

Incidental thoracic, hepatic and peritoneal calcifications: a case of Pentastomiasis

Incidental findings are not uncommon in radiology. In this case, although the incidental findings could be described as an Aunt Minnie, the patient underwent multiple investigations due to the rarity of the causative parasite. The current literature concerning Pentastomiasis suggests it may become more common in future. Our hope is that this case report will help future patients who present with the radiological pattern described to be more rapidly diagnosed and reassured.

Patients with inflammatory bowel disease have higher abdominal adiposity and less skeletal mass than healthy controls

Background

Abdominal fat type and distribution have been associated with complicated Crohn’s disease and adverse postoperative outcomes. Few studies have assessed the abdominal distribution of fat and lean stores in patients with inflammatory bowel disease (IBD) and compared this with healthy controls. This retrospective study aimed to compare the abdominal body composition in IBD patients who failed medical treatment and who underwent computed tomography (CT) imaging prior to gastrointestinal surgery with healthy controls. Associations between preoperative abdominal body composition and postoperative outcomes within a year of surgery were explored.

Methods

Abdominal body composition was evaluated in 22 presurgical patients with medically refractory IBD (18 with Crohn’s disease) and 22 healthy controls, using routinely acquired CT. Total fat, subcutaneous fat, visceral fat, and skeletal muscle cross-sectional area were measured.

Results

An independent disease effect was observed, explaining a fat deposition excess of 38 cm² and a skeletal muscle deficit of 15 cm² in IBD. Abdominal skeletal muscle correlated with visceral fat for the control (rho=0.51, P=0.015), but not for the IBD group (rho=-0.13, P=0.553). A positive correlation observed between subcutaneous fat with skeletal muscle in the controls (rho=0.47, P=0.026) was inverted in the IBD group (rho=-0.43, P=0.045). Preoperative abdominal body composition was not predictive of postoperative outcomes.

Conclusions

A higher degree of abdominal adiposity, a lower skeletal mass and a larger body size for the same anthropometry can be expected in IBD patients. Preoperative abdominal body composition is not associated with surgical outcomes.

Role of neoadjuvant treatment regimens for locally advanced pancreatic cancer

444

Background: Pancreatic ductal adenocarcinoma (PDAC) is a leading cause of cancer-related mortality worldwide. Lymph node involvement and resection margin status play important roles in predicting relapse. Resectable disease occurs in only 15–20% of total patients who present with PDAC. Unfortunately, margin involvement (R1) occurs in 70–80% of these patients. Emerging evidence has shown that the use of neoadjuvant chemotherapy and localised radiotherapy to downsize the tumours and increase the margin clearance (R0) rate may improve the overall survival of PDAC patients.We report a neoadjuvant therapy approach in the non-clinical trial setting of our large, tertiary cancer centre. Methods: We prospectively collected the outcome data and toxicity of 53 patients diagnosed with borderline resectable or initially non-resectable PDAC between 2012 and 2014. These patients received either FOLFIRINOX (FFX) or Gemcitabine/Capecitabine (GemCap) combination chemotherapies. Following restaging by computed tomography (CT), the patients proceeded to preo-operative 5-FU-based chemo-radiotherapy, immediate resection or subsequent palliativetherapies. Results: The median age was 65 (range 30 – 79) at PDAC diagnosis. Sixty-one percent (n=32) were male with the commonest anatomical location being the head of the pancreas (58%, n=31). The median follow up for survivors is 13.7 months (range: 5.3–24.4). The median overall survival was 18.3 months (95%, CI: 12.0–24.5). There was no statistical difference between overall survival in patients receiving FFX and GemCap chemotherapies. The margin clearance rate (R0) was 36% (4/11) in patients who proceeded to resection after neoadjuvant chemotherapy alone. The rate was 100% (4/4) in patients who received additional chemoradiation prior to surgery. Conclusions: This case series reveals that neoadjuvant therapy improved survival of patients with PDAC. In addition, we showed an increase in the R0 resection rate in patients who underwent chemoradiation prior to surgery. Further work is ongoing but based on historical data we believe that this neoadjuvant approach may lead to a long term survival benefit.