Development and clinical application of a deep learning model to identify acute infarct on magnetic resonance imaging

Stroke is a leading cause of death and disability. The ability to quickly identify the presence of acute infarct and quantify the volume on magnetic resonance imaging (MRI) has important treatment implications. We developed a machine learning model that used the apparent diffusion coefficient and diffusion weighted imaging series. It was trained on 6,657 MRI studies from Massachusetts General Hospital (MGH; Boston, USA). All studies were labelled positive or negative for infarct (classification annotation) with 377 having the region of interest outlined (segmentation annotation). The different annotation types facilitated training on more studies while not requiring the extensive time to manually segment every study. We initially validated the model on studies sequestered from the training set. We then tested the model on studies from three clinical scenarios: consecutive stroke team activations for 6-months at MGH, consecutive stroke team activations for 6-months at a hospital that did not provide training data (Brigham and Women’s Hospital [BWH]; Boston, USA), and an international site (Diagnósticos da América SA [DASA]; Brazil). The model results were compared to radiologist ground truth interpretations. The model performed better when trained on classification and segmentation annotations (area under the receiver operating curve [AUROC] 0.995 [95% CI 0.992–0.998] and median Dice coefficient for segmentation overlap of 0.797 [IQR 0.642–0.861]) compared to segmentation annotations alone (AUROC 0.982 [95% CI 0.972–0.990] and Dice coefficient 0.776 [IQR 0.584–0.857]). The model accurately identified infarcts for MGH stroke team activations (AUROC 0.964 [95% CI 0.943–0.982], 381 studies), BWH stroke team activations (AUROC 0.981 [95% CI 0.966–0.993], 247 studies), and at DASA (AUROC 0.998 [95% CI 0.993–1.000], 171 studies). The model accurately segmented infarcts with Pearson correlation comparing model output and ground truth volumes between 0.968 and 0.986 for the three scenarios. Acute infarct can be accurately detected and segmented on MRI in real-world clinical scenarios using a machine learning model.

HIV-infected CD4+ T Cells Use T-bet-dependent Pathway for Production of IL-10 Upon Antigen Recognition

Interleukin (IL)-10 has been implicated in persistence of pathogens in a number of chronic infections. Infected CD4+ cells upon reactivation with HIV antigens were also shown to produce IL-10, which might contribute to their persistence. Hence, it is crucial to determine mechanisms regulating IL-10 production after activation with HIV antigens for devising effective blocking strategies. In this study, ERK-, T-bet- and FoxP3-dependent pathways were evaluated for their possible roles in IL-10 production by infected CD4+ cells after reactivation with HIV Env. Intracellular and secreted IL-10 levels were determined by flow cytometry and Bioplex assay after treating PBMCs with PD98059, tipifarnib and cyclosporin A for blocking of ERK-, T-bet-and FoxP3-dependent pathways, respectively. Baseline levels of T-bet, pERK were higher in P24+ CD4+ cells as compared to uninfected CD4+ cells, which increased further after activation with Env. Inhibition of T-bet resulted in 2.3-fold reduction of IL-10 expression whereas ERK and FoxP3 inhibition failed to cause suppression of IL-10 expression. Conversely, IL-10 secreted by PBMCs was inhibited maximally after ERK inhibition suggesting its role in regulation of cytokine secretory pathway. IFN-γ was found to be suppressed after treatment with inhibitors of all these pathways. Thus, the study highlighted need for IL-10 blockade along with the use of antigens for therapeutic vaccinations or latency reversal and identified the T-bet-dependent pathway as an important pathway regulating IL-10 production by infected CD4+ cells. However, simultaneous blockade of IFN-γ precludes use of inhibitor of this pathway as an IL-10 blocking strategy.

Impact of GSTM1, GSTT1 and GSTP1 gene polymorphism and risk of ARV-associated hepatotoxicity in HIV-infected individuals and its modulation

Glutathione S-transferase (GST) family is involved in a two-stage detoxification process of a wide range of environmental toxins, carcinogen and antiretroviral (ARV) therapy (ART) drugs. The aim of this study is to describe the impact of genetic polymorphisms of GSTM1, GSTT1 and GSTP1-313A/G in the risk of ARV-associated hepatotoxicity in HIV-infected individuals and its modulation in hepatotoxic patients. We enrolled a total of 34 patients with hepatotoxicity, 131 HIV-infected individuals without hepatotoxicity under non-nucleoside reverse transcriptase inhibitor containing ART and 153 unrelated healthy individuals. With a case-control design, polymorphisms of GSTM1, GSTT1 and GSTP1-313A/G gene were genotyped by PCR and restriction enzyme-length polymorphism. Genotypes of GSTT1 null were significantly higher in HIV-infected individuals as compared with healthy controls (P=0.01, odds ratio (OR)=1.54). HIV-infected individuals with GSTM1-null genotype showed higher risk (P=0.09, OR=1.37) for hepatotoxicity, but risk was not significant. On evaluating gene-gene interaction models, GSTM1 null and GSTT1 null showed significant association with the risk of hepatotoxicity in HIV-infected individuals (P=0.004, OR=2.67) owing to synergistic effect of these genes. Individuals with GSTT1-null and GSTM1-null genotypes showed higher risk of hepatotoxicity with advanced stage of (CD4<200) of HIV infection (P=0.18, OR=1.39; P=0.63, OR=1.13). In case-only analysis, GSTT1-null genotype among alcohol users showed elevated risk of hepatotoxicity in HIV-infected individuals (P=0.12, OR=1.36, 95% confidence interval (CI): 0.94-1.97) as compared with GSTT1 genotypes. The carriers GSTM1-null+GSTT1-null genotype among nevirapine user showed prominent risk of hepatotoxicity in HIV-infected individuals (P=0.12, OR=4.21, 95% CI: 0.60-29.54). Hence, we can conclude that GSTT1-null and GSTM1-null genotypes alone and in combination may predict the acquisition of hepatotoxicity.

Nonoperative management of blunt splenic and hepatic trauma in the pediatric population: significant differences between adult and pediatric surgeons?

Although operative management was the preferred method of treating blunt abdominal trauma in the past, recent literature and practice recommend a nonsurgical approach to most pediatric splenic and hepatic injuries. The majority of data supporting the safety and efficacy of this nonoperative approach are derived from university trauma programs with a pediatric center where care was managed by pediatric surgeons only. To evaluate the applicability of this approach in a regional trauma center where pediatric patients are managed by pediatric and non-pediatric surgeons we reviewed the experience at a Level II community trauma center. Fifty-four children (16 years of age or less) were admitted between April 1992 and April 1998 after sustaining blunt traumatic splenic and/or hepatic injuries. There were 37 (69%) males and 17 (31%) females; the average age was 11 years (range 4 months to 16 years). Of the 54 patients 34 (63%) sustained splenic injuries, 17 (31%) sustained hepatic injuries, and three (6%) sustained both splenic and hepatic injuries. All of these injuries were diagnosed by CT scan or during laparotomy. The average Injury Severity Score was 14.9 with a range from four to 57. Of the 47 patients initially admitted for nonoperative management one patient failed nonoperative management and required operative intervention. In our study 98 per cent (46 of 47 patients) of pediatric patients were successfully managed nonoperatively. Complications of nonoperative management occurred in two patients. Both developed splenic pseudocysts after splenic injury, which required later operative repair. These data are comparable with those from university trauma programs and confirm that nonoperative management is safe in a community trauma center. The majority of children with blunt splenic and hepatic trauma can be successfully treated without surgery, in a regional trauma center treated by nonpediatric trauma surgeons, if the decision is based on careful initial evaluation, aggressive resuscitation, and close observation of their hemodynamic stability.

Cost savings associated with changes in routine laboratory tests ordered for victims of trauma

Not all trauma victims evaluated by the trauma service require a full complement of laboratory tests upon admission. This study set out to determine the cost savings and safety of limited laboratory testing of trauma victims. Before 1998, our admission trauma protocol included 11 laboratory tests for all trauma victims. In 1998, we created two categories: Trauma Blue--severe injury likely (Glasgow Coma Score <13; systolic blood pressure <100 mm Hg at any time; significant head, chest, abdominal, or proximal long bone injury; or clinical suspicion of need for operative or intensive care unit management) and Trauma Yellow--severe injury unlikely. The triage decision was made by the team leader or attending physician. Trauma Blue laboratory tests included an arterial blood gas, blood alcohol, type and screen or crossmatch, and urine dipstick. All patients who did not meet Trauma Blue criteria were entered in the Trauma Yellow group. There were only two tests for the Trauma Yellow group, a venous blood gas and blood alcohol. All arterial and venous blood gases measured pH, pO2, pCO2, HCO3, base deficit, hemoglobin, sodium, potassium, and ionized calcium. Other laboratory tests were done if requested by the trauma team leader or attending physician. All trauma admissions for a 3-month period were entered into this prospective study. The admitting trauma surgeon was surveyed after each admission to evaluate any problems in patient care. The test group was compared with a historical control of 100 consecutive patients under the original laboratory trauma protocol. One hundred and forty-eight (148) patients were entered into the study. Average laboratory cost per patient was $29.82 less with the study protocol. No patient care problem was identified. A cost savings of $29.82 per patient or $20,000.00 a year was realized for our institution, with no change in the quality of patient care. Trauma protocols designed to reflect a patient's potential for serious injury can result in a significant cost savings while preserving patient safety.