Deep Learning-Based Software Improves Clinicians' Detection Sensitivity of Aneurysms on Brain TOF-MRA

BACKGROUND AND PURPOSE

The detection of cerebral aneurysms on MRA is a challenging task. Recent studies have used deep learning-based software for automated detection of aneurysms on MRA and have reported high performance. The purpose of this study was to evaluate the incremental value of using deep learning-based software for the detection of aneurysms on MRA by 2 radiologists, a neurosurgeon, and a neurologist.

MATERIALS AND METHODS

TOF-MRA examinations of intracranial aneurysms were retrospectively extracted. Four physicians interpreted the MRA blindly. After a washout period, they interpreted MRA again using the software. Sensitivity and specificity per patient, sensitivity per lesion, and the number of false-positives per case were measured. Diagnostic performances, including subgroup analysis of lesions, were compared. Logistic regression with a generalized estimating equation was used.

RESULTS

A total of 332 patients were evaluated; 135 patients had positive findings with 169 lesions. With software assistance, patient-based sensitivity was statistically improved after the washout period (73.5% versus 86.5%, P < .001). The neurosurgeon and neurologist showed a significant increase in patient-based sensitivity with software assistance (74.8% versus 85.2%, P = .03, and 56.3% versus 84.4%, P < .001, respectively), while the number of false-positive cases did not increase significantly (23 versus 30, P = .20, and 22 versus 24, P = .75, respectively).

CONCLUSIONS

Software-aided reading showed significant incremental value in the sensitivity of clinicians in the detection of aneurysms on MRA without a significant increase in false-positive findings, especially for the neurosurgeon and neurologist. Software-aided reading showed equivocal value for the radiologist.

Development and Validation of a Deep Learning-Based Model to Distinguish Glioblastoma from Solitary Brain Metastasis Using Conventional MR Images

BACKGROUND AND PURPOSE

Differentiating glioblastoma from solitary brain metastasis preoperatively using conventional MR images is challenging. Deep learning models have shown promise in performing classification tasks. The diagnostic performance of a deep learning-based model in discriminating glioblastoma from solitary brain metastasis using preoperative conventional MR images was evaluated.

MATERIALS AND METHODS

Records of 598 patients with histologically confirmed glioblastoma or solitary brain metastasis at our institution between February 2006 and December 2017 were retrospectively reviewed. Preoperative contrast-enhanced T1WI and T2WI were preprocessed and roughly segmented with rectangular regions of interest. A deep neural network was trained and validated using MR images from 498 patients. The MR images of the remaining 100 were used as an internal test set. An additional 143 patients from another tertiary hospital were used as an external test set. The classifications of ResNet-50 and 2 neuroradiologists were compared for their accuracy, precision, recall, F1 score, and area under the curve.

RESULTS

The areas under the curve of ResNet-50 were 0.889 and 0.835 in the internal and external test sets, respectively. The area under the curve of neuroradiologists 1 and 2 were 0.889 and 0.768 in the internal test set and 0.857 and 0.708 in the external test set, respectively.

CONCLUSIONS

A deep learning-based model may be a supportive tool for preoperative discrimination between glioblastoma and solitary brain metastasis using conventional MR images.

BEAM or BuCyE high-dose chemotherapy followed by autologous stem cell transplantation in non-Hodgkin's lymphoma patients

7097

Background: The objective of this study was to compare the efficacy and toxicity of two high-dose regimens for autologous stem cell transplantation (ASCT) in patients with non-Hodgkin's lymphoma (NHL): BEAM (BCNU, etoposide, cytarabine, and melphalan) and BuCyE (busulfan, cyclophosphamide, and etoposide). Methods: We analysed 65 NHL patients, who underwent high-dose chemotherapy with BEAM (N=43) or BuCyE (N=22), followed by ASCT, at the Asan Medical Center. BEAM was used from February 2002 to October 2005, and BuCyE was used from November 2005 to April 2008. Results: Median age was 46 years (range: 15–68), and baseline characteristics, such as gender, International Prognostic Index (IPI), age adjusted IPI, stage and status of disease at ASCT, and median number of infused CD 34+cells/kg were well balanced between groups. The incidence of mucositis, nausea/vomiting, diarrhea and bleeding, and the number of events clinically important infections during ASCT did not differ between groups. Median follow-up for survivors was 49.3 months in the BEAM group and 21.5 months in the BuCyE group. Median overall survival (OS) was 30.6 months (95% confidence interval [CI], 8.19–53.0 months) and 22.6 months (95% CI, 12.1–33.1 months) and median event-free survival (EFS) was 16.1 months (95% CI, 0.0–53.6 months) and 11.2 months (95% CI, 0.0–22.5 months) in the BEAM and BuCyE group, respectively. There were no significant differences in OS (p=0.636) and EFS (p=0.575) between the two groups. Conclusions: In our analysis, BuCyE appeared to be not inferior to BEAM for survival. And we found that regimen-related toxicities did not differ significantly between the two groups.

No significant financial relationships to disclose.

The role of prophylactic antimicrobials during autologous stem cell transplantation: A single center experience

7105

Background: The aim of this retrospective study was to investigate the efficacy of antibiotic prophylaxis during autologous peripheral stem cell transplantation (ASCT) in patients with multiple myeloma (MM) and non-Hodgkin's lymphoma (NHL). Methods: We searched Asan Medical Center Registry for NHL and MM: Total 114 cases received antimicrobial prophylaxis; while 118 cases did not receive antimicrobial prophylaxis during ASCT. Results: In prophylaxis group, 80 of 114 (70.2%) patients had experienced febrile episodes at median day +6 after transplantation with a statistically significant difference (P<0.001). In no-prophylaxis group, 111 of 118 (94.1%) patients had experienced at median day +5. Documented infection occurred in 14 of 114 (12.3%) patients in prophylaxis group, and 16 of 118 (13.6%) patients in no-prophylaxis group (P=0.846). In these patients, the positive blood culture was seen in 12 (10.5%) of 114 patients in prophylaxis group, and 12 (10.7%) of 118 patients in no-prophylaxis group (P=1.000). Documented viral infection or reactivation was not observed in prophylaxis group, but observed in 4 patients of no-prophylaxis group. Both groups showed no invasive fungal infection or serious adverse event during ASCT. The day of infection resolved was a median day +15 (range, 3–29) in prophylaxis group and day +14 (range, 2–70) in no-prophylaxis group (P=0.945). The duration of antimicrobial treatment was median 10 days both in prophylaxis group and in no-prophylaxis group (P=0.565). Conclusions: In our experience, the antimicrobial prophylaxis seems to decrease the incidence of febrile episodes during ASCT, but seems to have no beneficial effect on reducing infectious complications. The antimicrobial prophylaxis of our study did not show the difference in the detection of causative organism as an infective agent, duration of antimicrobial therapy and hospitalization between two groups.

No significant financial relationships to disclose.

The prognostic significance of the number of extranodal sites in the patients with disseminated diffuse large B-cell lymphoma treated with R-CHOP

8570

Background: The combination of rituximab and CHOP chemotherapy (R-CHOP) has improved survival of patients with diffuse large B-cell lymphoma (DLBCL). Recently, several reports have shown that standard International Prognostic Index (IPI) became less powerful prognostic predictor in patients with DLBCL in the era of R-CHOP. We evaluated the prognostic factors of DLBCL patients treated with R-CHOP. Detailed analysis was planned regarding the number of extranodal sites because of its higher frequency in Korea. Methods: Between January 2002 and May 2008, 126 patients with stage III/IV DLBCL treated with R-CHOP were identified. We performed the retrospective analysis of the clinicopathologic factors and verified the predictive power of standard IPI and revised IPI (R-IPI) which was reported by the study group of British Columbia. Various numbers of extranodal sites were analyzed for further stratification and we set E-IPI as the IPI when the number of extranodal sites is stratified in ≤2 vs >2. Results: In the univariate analysis, the number of extranodal sites (≤2 vs >2) was a significant prognostic factor for complete response (CR) (p=0.04), event-free survival (EFS) (p=0.01) and overall survival (OS) (p<0.001). Age was also significant for EFS (p=0.03). When the number of extranodal site was stratified differently (0 vs >0, or ≤1 vs >1), these were not associated with CR, EFS and OS. On the multivariate analysis, the number of extranodal sites (≤2 vs >2) remained significant for EFS (p<0.01, HR 2.6) and OS (p<0.01, HR 3.5). The standard IPI identified 3 risk groups with 2-year EFS; 68%, 55%, 56% (p=0.17) and 2-year OS; 85%, 68%, 58%, respectively (p=0.04). The R-IPI classified 2 risk groups with 2-year EFS; 65%, 50% (p=0.02) and 2-year OS 76%, 62%, respectively (p=0.04). The E-IPI represented 3 risk groups with 2-year EFS; 79%, 56%, 42% (p=0.01) and 2-year OS; 86%, 70%, 39%, respectively (p=0.001). The patient group with survival of less than 50% was only recognized by E-IPI. Conclusions: The number of extranodal sites (≤2 vs >2) is the most significant prognostic factor of EFS and OS. Although all three indices remain predictive, E-IPI is the best model to identify the prognostic group in this cohort with stage III/IV DLBCL treated with R-CHOP.

No significant financial relationships to disclose.

Outcomes in patients with primary gastric diffuse large B-cell lymphoma after rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisolone (R-CHOP) chemotherapy

e19543

Background: The optimal therapy for primary gastric diffuse large B- cell lymphoma (DLBCL) still needs to be defined. The aim of this study was to investigate the patient's outcomes after rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisolone (R-CHOP) treatment in primary gastric DLBCL in a single institution. Methods: We searched AMC Registry for Non-Hodgkin's Lymphoma and found 26 patients with primary gastric DLBCL, who received R-CHOP as first-line chemotherapy. Ten of 26 patients had localized disease. Remaining patients had disseminated disease. R-CHOP was repeated every 21 days in all patients. Results: Overall, complete response (CR) was observed in 20 of 26 patients (76.9%). Three-year event free survival (EFS) and overall survival (OS) was 76.5% and 75.0%, respectively. After analyses of 10 patients with localized disease, we found that these patients had received a total 38 cycles, with a median of 3 cycles per patient. Of 10 patients, one patient had 2 cycles of R-CHOP, 4 had 3 cycles, and one had 4 cycles, all 6 patients above followed by consolidation radiotherapy. Remaining one patient and 4 patients had 5 cycles and 6 cycles of R-CHOP, respectively. In patients with localized disease, CR was observed in 10 of 10 patients (100%), and both 3-year EFS and OS was 100% (10 of 10 patients). In analyses with 16 patients with disseminated disease, all patients had received a total 91 cycles, with a median of 6 cycles per patient. In these patients, two patients had radiation therapy after R-CHOP, one patient had CR before consolidation radiation therapy, and another had partial response before radiation therapy. CR after R-CHOP treatment was observed in 10 of 16 patients (62.5%), partial response in 3 patients, stable disease in 1 patient, and progressive disease in 1 patient. Three-year EFS and OS was 61.1% and 57.8% in patients with disseminated disease. Conclusions: R-CHOP regimen showed a promising result in primary gastric DLBCL. Combination with rituximab in CHOP regimen showed excellent prognosis especially in patients with localized disease. In localized disease, CR was 100%, 3-year EFS and OS was 100%. In disseminated disease, CR was 62.5%, 3-year EFS and OS was 61.1% and 57.8%.

No significant financial relationships to disclose.