Purpose
To develop a head and neck normal structures autocontouring tool that could be used to automatically detect the errors in autocontours from a clinically validated autocontouring tool.
Methods
...An autocontouring tool based on convolutional neural networks (CNN) was developed for 16 normal structures of the head and neck and tested to identify the contour errors from a clinically validated multiatlas‐based autocontouring system (MACS). The computed tomography (CT) scans and clinical contours from 3495 patients were semiautomatically curated and used to train and validate the CNN‐based autocontouring tool. The final accuracy of the tool was evaluated by calculating the Sørensen–Dice similarity coefficients (DSC) and Hausdorff distances between the automatically generated contours and physician‐drawn contours on 174 internal and 24 external CT scans. Lastly, the CNN‐based tool was evaluated on 60 patients' CT scans to investigate the possibility to detect contouring failures. The contouring failures on these patients were classified as either minor or major errors. The criteria to detect contouring errors were determined by analyzing the DSC between the CNN‐ and MACS‐based contours under two independent scenarios: (a) contours with minor errors are clinically acceptable and (b) contours with minor errors are clinically unacceptable.
Results
The average DSC and Hausdorff distance of our CNN‐based tool was 98.4%/1.23 cm for brain, 89.1%/0.42 cm for eyes, 86.8%/1.28 cm for mandible, 86.4%/0.88 cm for brainstem, 83.4%/0.71 cm for spinal cord, 82.7%/1.37 cm for parotids, 80.7%/1.08 cm for esophagus, 71.7%/0.39 cm for lenses, 68.6%/0.72 for optic nerves, 66.4%/0.46 cm for cochleas, and 40.7%/0.96 cm for optic chiasm. With the error detection tool, the proportions of the clinically unacceptable MACS contours that were correctly detected were 0.99/0.80 on average except for the optic chiasm, when contours with minor errors are clinically acceptable/unacceptable, respectively. The proportions of the clinically acceptable MACS contours that were correctly detected were 0.81/0.60 on average except for the optic chiasm, when contours with minor errors are clinically acceptable/unacceptable, respectively.
Conclusion
Our CNN‐based autocontouring tool performed well on both the publically available and the internal datasets. Furthermore, our results show that CNN‐based algorithms are able to identify ill‐defined contours from a clinically validated and used multiatlas‐based autocontouring tool. Therefore, our CNN‐based tool can effectively perform automatic verification of MACS contours.
Mortality of patients with head and neck squamous cell carcinoma (HNSCC) is primarily driven by tumor cell radioresistance leading to locoregional recurrence (LRR). In this study, we use a ...classification of TP53 mutation (disruptive vs. nondisruptive) and examine impact on clinical outcomes and radiation sensitivity.
Seventy-four patients with HNSCC treated with surgery and postoperative radiation and 38 HNSCC cell lines were assembled; for each, TP53 was sequenced and the in vitro radioresistance measured using clonogenic assays. p53 protein expression was inhibited using short hairpin RNA (shRNA) and overexpressed using a retrovirus. Radiation-induced apoptosis, mitotic cell death, senescence, and reactive oxygen species (ROS) assays were carried out. The effect of the drug metformin on overcoming mutant p53-associated radiation resistance was examined in vitro as well as in vivo, using an orthotopic xenograft model.
Mutant TP53 alone was not predictive of LRR; however, disruptive TP53 mutation strongly predicted LRR (P = 0.03). Cell lines with disruptive mutations were significantly more radioresistant (P < 0.05). Expression of disruptive TP53 mutations significantly decreased radiation-induced senescence, as measured by SA-β-gal staining, p21 expression, and release of ROS. The mitochondrial agent metformin potentiated the effects of radiation in the presence of a disruptive TP53 mutation partially via senescence. Examination of our patient cohort showed that LRR was decreased in patients taking metformin.
Disruptive TP53 mutations in HNSCC tumors predicts for LRR, because of increased radioresistance via the inhibition of senescence. Metformin can serve as a radiosensitizer for HNSCC with disruptive TP53, presaging the possibility of personalizing HNSCC treatment.
To characterize long-term MD Anderson Dysphagia Inventory (MDADI) results after primary intensity modulated radiation therapy (IMRT) for oropharyngeal carcinoma (OPC) among patients with ..."low-intermediate risk" OPC who would be eligible for current trials (eg, ECOG 3311, NRG HN002, CRUK PATHOS).
A retrospective pooled analysis combined data from 3 single-institution clinical trials for advanced-stage head and neck carcinoma. Inclusion criteria were clinical stage III/IV OPC (T1-2/N1-2b, T3/N0-2b) treated with definitive split-field IMRT and prospectively collected MDADI at baseline and at least 1 posttreatment interval available in trial databases. Patients were sampled to represent likely human papillomavirus (HPV)-associated disease (HPV
/p16
or <10 pack-years if HPV/p16 unknown). The MDADI composite scores were collected at baseline and 6, 12, and 24 months after treatment. Pairwise tests were Bonferroni corrected for multiple comparisons.
Forty-six patients were included. All received bilateral neck irradiation with a median dose of 70 Gy and systemic therapy (57% concurrent, 43% induction only). Overall the mean baseline MDADI composite score was 90.1, dropping to 74.6 at 6 months (P<.0001) and rising to 78.5 (P<.0001) and 83.1 (P=.002) by 12 and 24 months relative to baseline, respectively, representing a clinically meaningful drop in MDADI scores at 6 months that partially recovers by 24 months (6 vs 24 months, P=.05). Poor MDADI scores (composite <60) were reported in 4%, 11%, 15%, and 9% of patients at baseline and 6, 12, and 24 months, respectively. Fifteen percent of patients had a persistently depressed composite score by at least 20 points at the 24-month interval.
"Low-intermediate risk" patients with OPC treated with laryngeal/esophageal inlet dose-optimized split-field IMRT are highly likely to report recovery of acceptable swallowing function in long-term follow-up. Only 15% report poor swallowing function and/or persistently depressed MDADI at 12 months or more after IMRT. These data serve as a benchmark future trial design and endpoint interpretation.
A single-institution prospective study was conducted to assess disease control and toxicity of proton therapy for patients with head and neck cancer.
Disease control, toxicity, functional outcomes, ...and patterns of failure for the initial cohort of patients with oropharyngeal squamous carcinoma (OPC) treated with intensity modulated proton therapy (IMPT) were prospectively collected in 2 registry studies at a single institution. Locoregional failures were analyzed by using deformable image registration.
Fifty patients with OPC treated from March 3, 2011, to July 2014 formed the cohort. Eighty-four percent were male, 50% had never smoked, 98% had stage III/IV disease, 64% received concurrent therapy, and 35% received induction chemotherapy. Forty-four of 45 tumors (98%) tested for p16 were positive. All patients received IMPT (multifield optimization to n=46; single-field optimization to n=4). No Common Terminology Criteria for Adverse Events grade 4 or 5 toxicities were observed. The most common grade 3 toxicities were acute mucositis in 58% of patients and late dysphagia in 12%. Eleven patients had a gastrostomy (feeding) tube placed during therapy, but none had a feeding tube at last follow-up. At a median follow-up time of 29 months, 5 patients had disease recurrence: local in 1, local and regional in 1, regional in 2, and distant in 1. The 2-year actuarial overall and progression-free survival rates were 94.5% and 88.6%.
The oncologic, toxicity, and functional outcomes after IMPT for OPC are encouraging and provide the basis for ongoing and future clinical studies.
To evaluate the magnitude of cervix regression and motion during external beam chemoradiation for cervical cancer.
Sixteen patients with cervical cancer underwent computed tomography scanning before, ...weekly during, and after conventional chemoradiation. Cervix volumes were calculated to determine the extent of cervix regression. Changes in the center of mass and perimeter of the cervix between scans were used to determine the magnitude of cervix motion. Maximum cervix position changes were calculated for each patient, and mean maximum changes were calculated for the group.
Mean cervical volumes before and after 45 Gy of external beam irradiation were 97.0 and 31.9 cc, respectively; mean volume reduction was 62.3%. Mean maximum changes in the center of mass of the cervix were 2.1, 1.6, and 0.82 cm in the superior-inferior, anterior-posterior, and right-left lateral dimensions, respectively. Mean maximum changes in the perimeter of the cervix were 2.3 and 1.3 cm in the superior and inferior, 1.7 and 1.8 cm in the anterior and posterior, and 0.76 and 0.94 cm in the right and left lateral directions, respectively.
Cervix regression and internal organ motion contribute to marked interfraction variations in the intrapelvic position of the cervical target in patients receiving chemoradiation for cervical cancer. Failure to take these variations into account during the application of highly conformal external beam radiation techniques poses a theoretical risk of underdosing the target or overdosing adjacent critical structures.
TP53 is the most frequently altered gene in head and neck squamous cell carcinoma, with mutations occurring in over two-thirds of cases, but the prognostic significance of these mutations remains ...elusive. In the current study, we evaluated a novel computational approach termed evolutionary action (EAp53) to stratify patients with tumors harboring TP53 mutations as high or low risk, and validated this system in both in vivo and in vitro models. Patients with high-risk TP53 mutations had the poorest survival outcomes and the shortest time to the development of distant metastases. Tumor cells expressing high-risk TP53 mutations were more invasive and tumorigenic and they exhibited a higher incidence of lung metastases. We also documented an association between the presence of high-risk mutations and decreased expression of TP53 target genes, highlighting key cellular pathways that are likely to be dysregulated by this subset of p53 mutations that confer particularly aggressive tumor behavior. Overall, our work validated EAp53 as a novel computational tool that may be useful in clinical prognosis of tumors harboring p53 mutations.
Squamous cell carcinoma driven by human papillomavirus (HPV) is more sensitive to DNA-damaging therapies than its HPV-negative counterpart. Here, we show that p16, the clinically used surrogate for ...HPV positivity, renders cells more sensitive to radiotherapy via a ubiquitin-dependent signaling pathway, linking high levels of this protein to increased activity of the transcription factor SP1, increased HUWE1 transcription, and degradation of ubiquitin-specific protease 7 (USP7) and TRIP12. Activation of this pathway in HPV-positive disease led to decreased homologous recombination and improved response to radiotherapy, a phenomenon that can be recapitulated in HPV-negative disease using USP7 inhibitors in clinical development. This p16-driven axis induced sensitivity to PARP inhibition and potentially leads to "BRCAness" in head and neck squamous cell carcinoma (HNSCC) cells. Thus, these findings support a functional role for p16 in HPV-positive tumors in driving response to DNA damage, which can be exploited to improve outcomes in both patients with HPV-positive and HPV-negative HNSCC.
In HPV-positive tumors, a previously undiscovered pathway directly links p16 to DNA damage repair and sensitivity to radiotherapy via a clinically relevant and pharmacologically targetable ubiquitin-mediated degradation pathway.
The future of artificial intelligence (AI) heralds unprecedented change for the field of radiation oncology. Commercial vendors and academic institutions have created AI tools for radiation oncology, ...but such tools have not yet been widely adopted into clinical practice. In addition, numerous discussions have prompted careful thoughts about AI's impact upon the future landscape of radiation oncology: How can we preserve innovation, creativity, and patient safety? When will AI-based tools be widely adopted into the clinic? Will the need for clinical staff be reduced? How will these devices and tools be developed and regulated?
In this work, we examine how deep learning, a rapidly emerging subset of AI, fits into the broader historical context of advancements made in radiation oncology and medical physics. In addition, we examine a representative set of deep learning-based tools that are being made available for use in external beam radiotherapy treatment planning and how these deep learning-based tools and other AI-based tools will impact members of the radiation treatment planning team. Key Messages: Compared to past transformative innovations explored in this article, such as the Monte Carlo method or intensity-modulated radiotherapy, the development and adoption of deep learning-based tools is occurring at faster rates and promises to transform practices of the radiation treatment planning team. However, accessibility to these tools will be determined by each clinic's access to the internet, web-based solutions, or high-performance computing hardware. As seen by the trends exhibited by many technologies, high dependence on new technology can result in harm should the product fail in an unexpected manner, be misused by the operator, or if the mitigation to an expected failure is not adequate. Thus, the need for developers and researchers to rigorously validate deep learning-based tools, for users to understand how to operate tools appropriately, and for professional bodies to develop guidelines for their use and maintenance is essential. Given that members of the radiation treatment planning team perform many tasks that are automatable, the use of deep learning-based tools, in combination with other automated treatment planning tools, may refocus tasks performed by the treatment planning team and may potentially reduce resource-related burdens for clinics with limited resources.