Semisupervised Training of a Brain MRI Tumor Detection Model Using Mined Annotations.

Background Artificial intelligence (AI) applications for cancer imaging conceptually begin with automated tumor detection, which can provide the foundation for downstream AI tasks. However, supervised training requires many image annotations, and performing dedicated post hoc image labeling is burdensome and costly. Purpose To investigate whether clinically generated image annotations can be data mined from the picture archiving and communication system (PACS), automatically curated, and used for semisupervised training of a brain MRI tumor detection model. Materials and Methods In this retrospective study, the cancer center PACS was mined for brain MRI scans acquired between January 2012 and December 2017 and included all annotated axial T1 postcontrast images. Line annotations were converted to boxes, excluding boxes shorter than 1 cm or longer than 7 cm. The resulting boxes were used for supervised training of object detection models using RetinaNet and Mask region-based convolutional neural network (R-CNN) architectures. The best-performing model trained from the mined data set was used to detect unannotated tumors on training images themselves (self-labeling), automatically correcting many of the missing labels. After self-labeling, new models were trained using this expanded data set. Models were scored for precision, recall, and F1 using a held-out test data set comprising 754 manually labeled images from 100 patients (403 intra-axial and 56 extra-axial enhancing tumors). Model F1 scores were compared using bootstrap resampling. Results The PACS query extracted 31 150 line annotations, yielding 11 880 boxes that met inclusion criteria. This mined data set was used to train models, yielding F1 scores of 0.886 for RetinaNet and 0.908 for Mask R-CNN. Self-labeling added 18 562 training boxes, improving model F1 scores to 0.935 (P < .001) and 0.954 (P < .001), respectively. Conclusion The application of semisupervised learning to mined image annotations significantly improved tumor detection performance, achieving an excellent F1 score of 0.954. This development pipeline can be extended for other imaging modalities, repurposing unused data silos to potentially enable automated tumor detection across radiologic modalities. © RSNA, 2022 Online supplemental material is available for this article.

Magnetic Resonance Imaging Screening for Trilateral Retinoblastoma: The Memorial Sloan Kettering Cancer Center Experience 2006-2016.

PURPOSE: Magnetic resonance imaging (MRI) has been used for baseline brain imaging and afterward as a screening tool for trilateral retinoblastoma (TRB), but there is no consensus on timing or frequency of screening worldwide. In this study, a cohort of hereditary retinoblastoma patients at increased risk for TRB was identified and the usefulness of aggressive neuroimaging was examined. DESIGN: Retrospective review of the medical records and MRI reports of patients with retinoblastoma treated at Memorial Sloan Kettering Cancer Center between January 1, 2006, and December 31, 2016. PARTICIPANTS: Three hundred forty-nine total patients with retinoblastoma, including 215 hereditary retinoblastoma patients in the screening group. METHODS: We reviewed 804 MRI studies of the orbit or brain. Patient and disease characteristics, including laterality, family history, and gene mutation status were analyzed. The impression of every MRI was coded 1 to 5, each value representing a different abnormality. MAIN OUTCOME MEASURES: We calculated the incidence of TRB in patients with germline disease as well as the incidence of screening MRI scans showing TRB. RESULTS: Among our hereditary retinoblastoma screening cohort (n=215) 4 patients with TRB were identified on screening MRI. All 4 patients showed bilateral disease, pineal gland tumors, and a latency period of at least 1 year. Three of the 4 were deceased by the end of the study. The incidence of TRB diagnosis was 1.9% (95% confidence interval [CI], 0.7%-4.9%). Of the 804 screening MRI scans performed on the screening cohort, 691 (86%) were unremarkable and 4 reported a lesion suspicious for TRB. The overall incidence of detecting TRB on screening MRI in the at-risk cohort was 0.5% (95% CI, 0.2%-1.3%) with a number needed to treat of 202. CONCLUSIONS: All cases of TRB in our center during the study period developed before the patient was 3 years of age and after a total of only 4 lifetime MRIs. Overall survival from TRB was not improved as a result of screening, and many false-positive results required additional, subsequent MRI scans with anesthesia.

A Phase 1b Study of Cetuximab and BYL719 (Alpelisib) Concurrent with Intensity Modulated Radiation Therapy in Stage III-IVB Head and Neck Squamous Cell Carcinoma.

PURPOSE: Activation of the PI3K/mTOR signaling pathway is common in head and neck squamous cell carcinoma (HNSCC). BYL719 is an α-specific PI3K inhibitor that is synergistic and efficacious when combined with cetuximab, a Food and Drug Administration-approved radiosensitizing agent in the treatment of HNSCC. The agent independently has been shown to enhance radiosensitivity. This study evaluates the addition of BYL719 to cetuximab and radiation in the treatment of locally advanced HNSCC. METHODS AND MATERIALS: This is a single-institution, phase 1 study. Patients with American Joint Committee on Cancer seventh edition stage III to IVB HNSCC received standard cetuximab (400 mg/m2 intravenous loading dose) before intensity modulated radiation therapy (IMRT) followed by 250 mg/m2 weekly infusions during IMRT. BYL719 was given orally during IMRT in 3 dose levels: (1) 200 mg/d, (2) 250 mg/d, or (3) 300 mg/d in a standard 3 + 3 dose-escalation design. RESULTS: Eleven patients were evaluable. Dose level 2 was the maximum tolerated dose for BYL719. Two patients on dose level 3 had dose-limiting toxicities of oral mucositis that required a dose reduction of BYL719. One patient on dose level 2 had a dose-limiting toxicity of nausea that led to withdrawal of on-study treatment. Related grade 3 or higher adverse events consisted of decreased lymphocyte count, oral mucositis, dysphagia, hyperglycemia, maculopapular rash, and palmar-plantar erythrodysesthesia syndrome. All 11 patients had a complete response on posttreatment imaging, and 10 remain disease free. Of the 8 patients with mutational analysis, 1 had an activating PIK3CA mutation associated with a rapid response on serial intratreatment magnetic resonance imaging scans. CONCLUSIONS: The recommended phase 2 dose of BYL719 is 250 mg/d in combination with cetuximab and IMRT in patients with locally advanced HNSCC. Further evaluation of the addition of BYL719 to the platinum-sparing regimen of cetuximab and IMRT in the treatment of locally advanced HNSCC is warranted.

Selumetinib in paediatric patients with BRAF-aberrant or neurofibromatosis type 1-associated recurrent, refractory, or progressive low-grade glioma: a multicentre, phase 2 trial.

BACKGROUND: Paediatric low-grade glioma is the most common CNS tumour of childhood. Although overall survival is good, disease often recurs. No single universally accepted treatment exists for these patients; however, standard cytotoxic chemotherapies are generally used. We aimed to assess the activity of selumetinib, a MEK1/2 inhibitor, in these patients. METHODS: The Pediatric Brain Tumor Consortium performed a multicentre, phase 2 study in patients with paediatric low-grade glioma in 11 hospitals in the USA. Patients aged 3-21 years with a Lansky or Karnofsky performance score greater than 60 and the presence of recurrent, refractory, or progressive paediatric low-grade glioma after at least one standard therapy were eligible for inclusion. Patients were assigned to six unique strata according to histology, tumour location, NF1 status, and BRAF aberration status; herein, we report the results of strata 1 and 3. Stratum 1 comprised patients with WHO grade I pilocytic astrocytoma harbouring either one of the two most common BRAF aberrations (KIAA1549-BRAF fusion or the BRAFV600E [Val600Glu] mutation). Stratum 3 comprised patients with any neurofibromatosis type 1 (NF1)-associated paediatric low-grade glioma (WHO grades I and II). Selumetinib was provided as capsules given orally at the recommended phase 2 dose of 25 mg/m2 twice daily in 28-day courses for up to 26 courses. The primary endpoint was the proportion of patients with a stratum-specific objective response (partial response or complete response), as assessed by the local site and sustained for at least 8 weeks. All responses were reviewed centrally. All eligible patients who initiated treatment were evaluable for the activity and toxicity analyses. Although the trial is ongoing in other strata, enrolment and planned follow-up is complete for strata 1 and 3. This trial is registered with ClinicalTrials.gov, number NCT01089101. FINDINGS: Between July 25, 2013, and June 12, 2015, 25 eligible and evaluable patients were accrued to stratum 1, and between Aug 28, 2013, and June 25, 2015, 25 eligible and evaluable patients were accrued to stratum 3. In stratum 1, nine (36% [95% CI 18-57]) of 25 patients achieved a sustained partial response. The median follow-up for the 11 patients who had not had a progression event by Aug 9, 2018, was 36·40 months (IQR 21·72-45·59). In stratum 3, ten (40% [21-61]) of 25 patients achieved a sustained partial response; median follow-up was 48·60 months (IQR 39·14-51·31) for the 17 patients without a progression event by Aug 9, 2018. The most frequent grade 3 or worse adverse events were elevated creatine phosphokinase (five [10%]) and maculopapular rash (five [10%]). No treatment-realted deaths were reported. INTERPRETATION: Selumetinib is active in recurrent, refractory, or progressive pilocytic astrocytoma harbouring common BRAF aberrations and NF1-associated paediatric low-grade glioma. These results show that selumetinib could be an alternative to standard chemotherapy for these subgroups of patients, and have directly led to the development of two Children's Oncology Group phase 3 studies comparing standard chemotherapy to selumetinib in patients with newly diagnosed paediatric low-grade glioma both with and without NF1. FUNDING: National Cancer Institute Cancer Therapy Evaluation Program, the American Lebanese Syrian Associated Charities, and AstraZeneca.

Phase 2 study of vascular endothelial growth factor trap for the treatment of metastatic thyroid cancer.

BACKGROUND: Several multitargeted tyrosine kinase inhibitors (TKIs) have demonstrated activity in patients with thyroid cancer that is refractory to radioactive iodine (RAI). The antitumor effect is attributed at least in part to the ability of these TKIs to inhibit angiogenesis in these vascular tumors. Vascular endothelial growth factor (VEGF) Trap (VT) is a recombinantly produced fusion protein consisting solely of human sequences for VEGF receptors 1 and 2 extracellular domains and human immunoglobulin 1. Evaluating VT in patients with thyroid cancer is reasonable considering the activity observed with TKIs targeting VEGF. METHODS: The current study was a single-institution, phase 2, Simon 2-stage design (21 to >41 patients) study based on the objective response rate and/or 6-month progression-free survival as the primary endpoints. Eligible patients were required to have progressive, RAI-refractory and/or [18 F]fludeoxyglucose-avid, recurrent and/or metastatic, nonmedullary, nonanaplastic thyroid cancer; disease that was measurable using Response Evaluation Criteria In Solid Tumors (RECIST) criteria; and adequate organ and bone marrow function. VT at a dose of 4 mg/kg intravenously was administered every 14 days. RESULTS: A total of 40 patients were included in the analysis. Of these patients, 24 had papillary thyroid cancer, 2 had follicular thyroid cancer, and 11 had Hurthle cell thyroid cancer. The final 3 tumors were classified as poorly differentiated. There were no complete and/or partial responses noted; 34 patients achieved stable disease and 6 patients experienced disease progression as their best response. Of the 34 patients with stable disease, 16 remained on the study for >6 months and 6 patients remained on the study for >12 months. The median duration on treatment was 4.1 months (range, 0.6-30.8 months). CONCLUSIONS: Unlike TKIs, which have shown responses in this setting, to the authors' knowledge there have been no responses observed with the use of single-agent VT to date. It does not appear to be a promising drug for the treatment of patients with thyroid cancer.

Phase II Study of Lenvatinib in Patients With Progressive, Recurrent or Metastatic Adenoid Cystic Carcinoma.

PURPOSE: Recurrent or metastatic adenoid cystic carcinoma (R/M ACC) is a malignant neoplasm of predominantly salivary gland origin for which effective therapies are lacking. We conducted a phase II trial evaluating the multitargeted tyrosine kinase inhibitor lenvatinib in patients with R/M ACC. PATIENTS AND METHODS: This study was conducted with a two-stage minimax design. Patients with histologically confirmed R/M ACC of any primary site with radiographic and/or symptomatic progression were eligible. Any prior therapy was allowed except previous lenvatinib. Patients received lenvatinib 24 mg orally per day. The primary end point was overall response rate. Secondary end points were progression-free survival and safety. An exploratory analysis of how MYB expression and genomic alterations relate to outcomes was conducted. RESULTS: Thirty-three patients were enrolled; 32 were evaluable for the primary end point. Five patients (15.6%) had a confirmed partial response, 24 patients (75%) had stable disease, two patients (6.3%) discontinued treatment as a result of toxicity before the first scan, and one patient (3.1%) had progression of disease as best response. Median progression-free survival time was 17.5 months (95% CI, 7.2 months to not reached), although only eight progression events were observed. Patients otherwise were removed for toxicity (n = 5), as a result of withdrawal of consent (n = 9), or at the treating physician's discretion (n = 6). Twenty-three patients required at least one dose modification, and 18 of 32 patients discontinued lenvatinib for drug-related issues. The most common grade 3 or 4 adverse events were hypertension (n = 9; 28.1%) and oral pain (n = 3; 9.4%). Three grade 4 adverse events were observed (myocardial infarction, n = 1; posterior reversible encephalopathy syndrome, n = 1; and intracranial hemorrhage, n = 1). CONCLUSION: This trial met the prespecified overall response rate primary end point, demonstrating antitumor activity with lenvatinib in R/M ACC patients. Toxicity was comparable to previous studies, requiring monitoring and management.

Phase I study of induction chemotherapy with afatinib, ribavirin, and weekly carboplatin and paclitaxel for stage IVA/IVB human papillomavirus-associated oropharyngeal squamous cell cancer.

BACKGROUND: The human papillomavirus (HPV) E6 oncoprotein enhances the oncogenic potential of ErbB proteins in HPV-related malignancies. This phase I study evaluates the addition of afatinib, an ErbB family inhibitor, and ribavirin to paclitaxel and carboplatin induction chemotherapy in HPV-associated, locally advanced oropharyngeal squamous cell carcinoma (SCC). METHODS: This dose escalation study included 2 doses of oral afatinib: 30 and 40 mg daily. Ribavirin dosing was weight based. Paclitaxel (80 mg/m2 ) and carboplatin (area under the curve [AUC] 1.5) were administered on days 1 and 8 of each 21-day cycle. After 3 cycles, patients were removed from protocol to receive definitive treatment. RESULTS: Among 10 patients, there were no dose-limiting toxicities. Six patients (67%) had unconfirmed objective partial responses. The 2-year progression-free survival rate was 75%. CONCLUSION: Afatinib, ribavirin, paclitaxel, and carboplatin induction chemotherapy is safe and well tolerated. The phase II recommended dose of afatinib is 40 mg oral daily in this combination regimen.

Extraneural ependymoma: distant bone, lung, liver, and lymph node metastases following bevacizumab.

Extraneural metastases of ependymoma are rare, and have been reported in the lungs, lymph nodes, pleura, mediastinum, liver, diaphragmatic muscle, and bone. We report a case of anaplastic ependymoma with distant metastases to the vertebral bones, lungs, liver, and lymph nodes following treatment with bevacizumab. Recent research has hypothesized that angiogenic tumors may develop means of resistance to antiangiogenic therapies, and some evidence suggests potential for antiangiogenic therapies to promote additional means for cancer spread. Nevertheless, antiangiogenic therapies continue to demonstrate potential as potent therapies for the treatment of many cancers, and should continue to be researched for future uses.