Efficient and unique cobarcoding of second-generation sequencing reads from long DNA molecules enabling cost-effective and accurate sequencing, haplotyping, and de novo assembly.

Here, we describe single-tube long fragment read (stLFR), a technology that enables sequencing of data from long DNA molecules using economical second-generation sequencing technology. It is based on adding the same barcode sequence to subfragments of the original long DNA molecule (DNA cobarcoding). To achieve this efficiently, stLFR uses the surface of microbeads to create millions of miniaturized barcoding reactions in a single tube. Using a combinatorial process, up to 3.6 billion unique barcode sequences were generated on beads, enabling practically nonredundant cobarcoding with 50 million barcodes per sample. Using stLFR, we demonstrate efficient unique cobarcoding of more than 8 million 20- to 300-kb genomic DNA fragments. Analysis of the human genome NA12878 with stLFR demonstrated high-quality variant calling and phase block lengths up to N50 34 Mb. We also demonstrate detection of complex structural variants and complete diploid de novo assembly of NA12878. These analyses were all performed using single stLFR libraries, and their construction did not significantly add to the time or cost of whole-genome sequencing (WGS) library preparation. stLFR represents an easily automatable solution that enables high-quality sequencing, phasing, SV detection, scaffolding, cost-effective diploid de novo genome assembly, and other long DNA sequencing applications.

Trends in Adjuvant External Beam Radiation Therapy for Nonanaplastic Thyroid Cancer in California, 2003-2017.

BACKGROUND: External beam radiation therapy (EBRT) is rarely used to treat patients with differentiated or medullary thyroid cancer. Although EBRT is generally administered to patients with high-risk or unresectable diseases, neither its indications for the use nor the associated outcomes are well-defined. We used a statewide cohort to assess the trends in EBRT use and postradiation outcomes in California. METHODS: A population-based study of patients within the California Cancer Registry who underwent EBRT after surgery for nonanaplastic thyroid cancer (2003-2017) was conducted. The primary outcome was the annual utilization rate of EBRT. The secondary outcomes included Kaplan-Meier analysis for cause-specific survival and identifying factors associated with improved survival after EBRT. RESULTS: Among the 57 607 patients with nonanaplastic thyroid cancer from 2003 to 2017, 344 (0.6%) patients received EBRT. EBRT was utilized in 0.4% of papillary, 1.1% of follicular, and 7.7% of medullary thyroid cancers in California. Overall, 99 (28.8%) patients treated with EBRT died of thyroid cancer. The 10-year cause-specific survival of all patients with thyroid cancer after EBRT was 61.5% (95% CI: 54.8%-69.1%) and that of patients without distant disease was 80.3% (95% CI: 73.5%-87.8%). The survival outcomes varied by tumor size, histology, disease stage, patient age at diagnosis, and the presence of extrathyroidal extension (P < .05). CONCLUSIONS: The use of adjuvant EBRT for nonanaplastic thyroid cancer remained stable and low in California from 2003 to 2017. The comparative efficacy of EBRT was not discernible in this study, but disease control appeared durable in select patients. Well-controlled observational studies and/or prospective studies are needed to better define which patients benefit from EBRT.

NCCN Guidelines® Insights: Squamous Cell Skin Cancer, Version 1.2022.

The NCCN Guidelines for Squamous Cell Skin Cancer provide recommendations for diagnostic workup, clinical stage, and treatment options for patients with cutaneous squamous cell carcinoma. The NCCN panel meets annually to discuss updates to the guidelines based on comments from panel members and the Institutional Review, as well as submissions from within NCCN and external organizations. These NCCN Guidelines Insights focus on the introduction of a new surgical recommendation terminology (peripheral and deep en face margin assessment), as well as recent updates on topical prophylaxis, immunotherapy for regional and metastatic disease, and radiation therapy.

Clinical assessment of geometric distortion for a 0.35T MR-guided radiotherapy system.

PURPOSE: To estimate the overall spatial distortion on clinical patient images for a 0.35 T MR-guided radiotherapy system. METHODS: Ten patients with head-and-neck cancer underwent CT and MR simulations with identical immobilization. The MR images underwent the standard systematic distortion correction post-processing. The images were rigidly registered and landmark-based analysis was performed by an anatomical expert. Distortion was quantified using Euclidean distance between each landmark pair and tagged by tissue interface: bone-tissue, soft tissue, or air-tissue. For baseline comparisons, an anthropomorphic phantom was imaged and analyzed. RESULTS: The average spatial discrepancy between CT and MR landmarks was 1.15 ± 1.14 mm for the phantom and 1.46 ± 1.78 mm for patients. The error histogram peaked at 0-1 mm. 66% of the discrepancies were <2 mm and 51% <1 mm. In the patient data, statistically significant differences (p-values < 0.0001) were found between the different tissue interfaces with averages of 0.88 ± 1.24 mm, 2.01 ± 2.20 mm, and 1.41 ± 1.56 mm for the air/tissue, bone/tissue, and soft tissue, respectively. The distortion generally correlated with the in-plane radial distance from the image center along the longitudinal axis of the MR. CONCLUSION: Spatial distortion remains in the MR images after systematic distortion corrections. Although the average errors were relatively small, large distortions observed at bone/tissue interfaces emphasize the need for quantitative methods for assessing and correcting patient-specific spatial distortions.

Clinical and genetic analysis of a rare syndrome associated with neoteny.

PurposeWe describe a novel syndrome in seven female patients with extreme developmental delay and neoteny.MethodsAll patients in this study were female, aged 4 to 23 years, were well below the fifth percentile in height and weight, had failed to develop sexually, and lacked the use of language. Karyotype and array chromosome genomic hybridization analysis failed to identify large-scale structural variations. To further understand the underlying cause of disease in these patients, whole-genome sequencing was performed.ResultsIn five patients, coding de novo mutations (DNMs) were found in five different genes. These genes fell into similar functional categories of transcription regulation and chromatin modification. Comparison to a control population suggested that individuals with neotenic complex syndrome (NCS)-a name that we propose herein-could have an excess of rare inherited variants in genes associated with developmental delay and autism, although the difference was not significant.ConclusionWe describe an extreme form of developmental delay, with the defining characteristic of neoteny. In most patients we identified coding DNMs in a set of genes intolerant of haploinsufficiency; however, it is not clear whether these contributed to NCS. Rare inherited variants may also be associated with NCS, but more samples need to be analyzed to achieve statistical significance.

Advanced Whole-Genome Sequencing and Analysis of Fetal Genomes from Amniotic Fluid.

BACKGROUND: Amniocentesis is a common procedure, the primary purpose of which is to collect cells from the fetus to allow testing for abnormal chromosomes, altered chromosomal copy number, or a small number of genes that have small single- to multibase defects. Here we demonstrate the feasibility of generating an accurate whole-genome sequence of a fetus from either the cellular or cell-free DNA (cfDNA) of an amniotic sample. METHODS: cfDNA and DNA isolated from the cell pellet of 31 amniocenteses were sequenced to approximately 50× genome coverage by use of the Complete Genomics nanoarray platform. In a subset of the samples, long fragment read libraries were generated from DNA isolated from cells and sequenced to approximately 100× genome coverage. RESULTS: Concordance of variant calls between the 2 DNA sources and with parental libraries was >96%. Two fetal genomes were found to harbor potentially detrimental variants in chromodomain helicase DNA binding protein 8 (CHD8) and LDL receptor-related protein 1 (LRP1), variations of which have been associated with autism spectrum disorder and keratosis pilaris atrophicans, respectively. We also discovered drug sensitivities and carrier information of fetuses for a variety of diseases. CONCLUSIONS: We were able to elucidate the complete genome sequence of 31 fetuses from amniotic fluid and demonstrate that the cfDNA or DNA from the cell pellet can be analyzed with little difference in quality. We believe that current technologies could analyze this material in a highly accurate and complete manner and that analyses like these should be considered for addition to current amniocentesis procedures.

Detecting clinically actionable variants in the 3' exons of PMS2 via a reflex workflow based on equivalent hybrid capture of the gene and its pseudogene.

BACKGROUND: Hereditary cancer screening (HCS) for germline variants in the 3' exons of PMS2, a mismatch repair gene implicated in Lynch syndrome, is technically challenging due to homology with its pseudogene PMS2CL. Sequences of PMS2 and PMS2CL are so similar that next-generation sequencing (NGS) of short fragments-common practice in multigene HCS panels-may identify the presence of a variant but fail to disambiguate whether its origin is the gene or the pseudogene. Molecular approaches utilizing longer DNA fragments, such as long-range PCR (LR-PCR), can definitively localize variants in PMS2, yet applying such testing to all samples can have logistical and economic drawbacks. METHODS: To address these drawbacks, we propose and characterize a reflex workflow for variant discovery in the 3' exons of PMS2. We cataloged the natural variation in PMS2 and PMS2CL in 707 samples and designed hybrid-capture probes to enrich the gene and pseudogene with equal efficiency. For PMS2 exon 11, NGS reads were aligned, filtered using gene-specific variants, and subject to standard diploid variant calling. For PMS2 exons 12-15, the NGS reads were permissively aligned to PMS2, and variant calling was performed with the expectation of observing four alleles (i.e., tetraploid calling). In this reflex workflow, short-read NGS identifies potentially reportable variants that are then subject to disambiguation via LR-PCR-based testing. RESULTS: Applying short-read NGS screening to 299 HCS samples and cell lines demonstrated >99% analytical sensitivity and >99% analytical specificity for single-nucleotide variants (SNVs) and short insertions and deletions (indels), as well as >96% analytical sensitivity and >99% analytical specificity for copy-number variants. Importantly, 92% of samples had resolved genotypes from short-read NGS alone, with the remaining 8% requiring LR-PCR reflex. CONCLUSION: Our reflex workflow mitigates the challenges of screening in PMS2 and serves as a guide for clinical laboratories performing multigene HCS. To facilitate future exploration and testing of PMS2 variants, we share the raw and processed LR-PCR data from commercially available cell lines, as well as variant frequencies from a diverse patient cohort.

Prognostic significance of HPV status in the re-irradiation of recurrent and second primary cancers of the head and neck.

PURPOSE: To evaluate the prognostic significance of human papillomavirus (HPV) status among patients treated by salvage radiation therapy for local-regional recurrences and second primary cancers of the head and neck arising in a previously irradiated field. METHODS AND MATERIALS: The medical records of 54 consecutive patients who underwent re-irradiation for squamous cell carcinoma of the head and neck occurring in a previously irradiated field were reviewed. Only patients with biopsy-proven evidence of recurrent disease that had previously been treated with doses of radiation therapy of at least 60 Gy were included. Determination of HPV status at the time of recurrence was performed by p16 immunohistochemistry. The median age at re-irradiation was 58.5 years (range, 27.9 to 81.5 years). Thirty patients (55.5%) were lifelong never-smokers. The Kaplan Meier method was used to calculate overall survival, progression-free survival, and local-regional control, and distant metastasis-free survival with comparisons between groups performed using the log-rank test. RESULTS: HPV status among tumors that were re-irradiated was as follows: 16 positive (29.7%); 7 negative (12.9%); 31 unknown (57.4%). The median overall survival in the entire cohort was 11.7 months (range, 8 to 27 months), with the 1-year and 2-year estimates of overall survival being 47.2% and 38.4%, respectively. A statistical trend was identified favoring patients with HPV-positive cancers with respect to the endpoints of overall survival (p = 0.06) and progression-free survival (p = 0.08) after re-irradiation when compared to the HPV-negative/unknown population. There was no significant difference in distant control between the two cohorts (p = 0.40). CONCLUSIONS: The favorable prognostic significance of HPV seemingly extends to patients treated by re-irradiation suggesting that this biomarker may be useful in risk stratification in this setting.

Prognostic significance of p16 in squamous cell carcinoma of the larynx and hypopharynx.

PURPOSE: To evaluate the prognostic significance of p16 expression among patients with squamous cell carcinoma of the larynx (LSCC) and hypopharynx (HSCC). METHODS: The medical records of all patients with locally advanced, non-metastatic LSCC/HSCC were reviewed. p16INK4A (p16) protein expression was evaluated on pathological specimens by immunohistochemistry (IHC), and the Kaplan-Meier method was used to estimate overall survival (OS) and locoregional control (LRC). In select cases, p16 expression was correlated to high-risk and low-risk HPV genotypes using in situ hybridization (ISH). RESULTS: Thirty-one patients (23 LSCC; 8 HSCC) were identified. Seventeen (54.8%) patients were p16 negative; 14 (45.2%) were p16-positive. The primary treatment modality was radiation therapy for 22 (71.0%) patients and surgery for 9 (29.0%). Nineteen (61.3%) patients were evaluated for high-risk HPV and low-risk HPV genotypes by IHC, of whom 2 (10.5%) patients were positive for high-risk HPV and 1 (5.3%) was positive for low-risk HPV. For high-risk HPV, the positive predictive value (PPV), sensitivity, and specificity of p16 was 20.0%, 100%, and 52.9%. There was no significant difference in the 2-year actuarial rates of OS (91% vs. 64%, p=0.34) or LRC (51% vs. 46%, p=0.69) between the p16-positive and p-16 negative patients. CONCLUSION: In this small cohort of 31 LSCC and HSCC patients, p16 was not a significant predictive of either LRC or OS. Furthermore, p16 was poorly correlated with HPV genotyping as identified by ISH.

SREBP-2-deficient and hypomorphic mice reveal roles for SREBP-2 in embryonic development and SREBP-1c expression.

Cholesterol and fatty acid biosynthesis are regulated by the sterol regulatory element-binding proteins (SREBPs), encoded by Srebf1 and Srebf2. We generated mice that were either deficient or hypomorphic for SREBP-2. SREBP-2 deficiency generally caused death during embryonic development. Analyses of Srebf2(-/-) embryos revealed a requirement for SREBP-2 in limb development and expression of morphogenic genes. We encountered only one viable Srebf2(-/-) mouse, which displayed alopecia, attenuated growth, and reduced adipose tissue stores. Hypomorphic SREBP-2 mice (expressing low levels of SREBP-2) survived development, but the female mice exhibited reduced body weight and died between 8 and 12 weeks of age. Male hypomorphic mice were viable but had reduced cholesterol stores in the liver and lower expression of SREBP target genes. Reduced SREBP-2 expression affected SREBP-1 isoforms in a tissue-specific manner. In the liver, reduced SREBP-2 expression nearly abolished Srebf1c transcripts and reduced Srebf1a mRNA levels. In contrast, adipose tissue displayed normal expression of SREBP target genes, likely due to a compensatory increase in Srebf1a expression. Our results establish that SREBP-2 is critical for survival and limb patterning during development. Reduced expression of SREBP-2 from the hypomorphic allele leads to early death in females and reduced cholesterol content in the liver, but not in adipose tissue.

Three differentiation states risk-stratify bladder cancer into distinct subtypes.

Current clinical judgment in bladder cancer (BC) relies primarily on pathological stage and grade. We investigated whether a molecular classification of tumor cell differentiation, based on a developmental biology approach, can provide additional prognostic information. Exploiting large preexisting gene-expression databases, we developed a biologically supervised computational model to predict markers that correspond with BC differentiation. To provide mechanistic insight, we assessed relative tumorigenicity and differentiation potential via xenotransplantation. We then correlated the prognostic utility of the identified markers to outcomes within gene expression and formalin-fixed paraffin-embedded (FFPE) tissue datasets. Our data indicate that BC can be subclassified into three subtypes, on the basis of their differentiation states: basal, intermediate, and differentiated, where only the most primitive tumor cell subpopulation within each subtype is capable of generating xenograft tumors and recapitulating downstream populations. We found that keratin 14 (KRT14) marks the most primitive differentiation state that precedes KRT5 and KRT20 expression. Furthermore, KRT14 expression is consistently associated with worse prognosis in both univariate and multivariate analyses. We identify here three distinct BC subtypes on the basis of their differentiation states, each harboring a unique tumor-initiating population.

The CD47-signal regulatory protein alpha (SIRPa) interaction is a therapeutic target for human solid tumors.

CD47, a "don't eat me" signal for phagocytic cells, is expressed on the surface of all human solid tumor cells. Analysis of patient tumor and matched adjacent normal (nontumor) tissue revealed that CD47 is overexpressed on cancer cells. CD47 mRNA expression levels correlated with a decreased probability of survival for multiple types of cancer. CD47 is a ligand for SIRPα, a protein expressed on macrophages and dendritic cells. In vitro, blockade of CD47 signaling using targeted monoclonal antibodies enabled macrophage phagocytosis of tumor cells that were otherwise protected. Administration of anti-CD47 antibodies inhibited tumor growth in orthotopic immunodeficient mouse xenotransplantation models established with patient tumor cells and increased the survival of the mice over time. Anti-CD47 antibody therapy initiated on larger tumors inhibited tumor growth and prevented or treated metastasis, but initiation of the therapy on smaller tumors was potentially curative. The safety and efficacy of targeting CD47 was further tested and validated in immune competent hosts using an orthotopic mouse breast cancer model. These results suggest all human solid tumor cells require CD47 expression to suppress phagocytic innate immune surveillance and elimination. These data, taken together with similar findings with other human neoplasms, show that CD47 is a commonly expressed molecule on all cancers, its function to block phagocytosis is known, and blockade of its function leads to tumor cell phagocytosis and elimination. CD47 is therefore a validated target for cancer therapies.

MUsculo-Skeleton-Aware (MUSA) deep learning for anatomically guided head-and-neck CT deformable registration.

Deep-learning-based deformable image registration (DL-DIR) has demonstrated improved accuracy compared to time-consuming non-DL methods across various anatomical sites. However, DL-DIR is still challenging in heterogeneous tissue regions with large deformation. In fact, several state-of-the-art DL-DIR methods fail to capture the large, anatomically plausible deformation when tested on head-and-neck computed tomography (CT) images. These results allude to the possibility that such complex head-and-neck deformation may be beyond the capacity of a single network structure or a homogeneous smoothness regularization. To address the challenge of combined multi-scale musculoskeletal motion and soft tissue deformation in the head-and-neck region, we propose a MUsculo-Skeleton-Aware (MUSA) framework to anatomically guide DL-DIR by leveraging the explicit multiresolution strategy and the inhomogeneous deformation constraints between the bony structures and soft tissue. The proposed method decomposes the complex deformation into a bulk posture change and residual fine deformation. It can accommodate both inter- and intra- subject registration. Our results show that the MUSA framework can consistently improve registration accuracy and, more importantly, the plausibility of deformation for various network architectures. The code will be publicly available at https://github.com/HengjieLiu/DIR-MUSA.

Multifunctional Surface Treatment against Imperfections and Halide Segregation in Wide-Band Gap Perovskite Solar Cells.

Mixed-halide wide-band gap perovskites (WBPs) still suffer from losses due to imperfections within the absorber and the segregation of halide ions under external stimuli. Herein, we design a multifunctional passivator (MFP) by mixing bromide salt, formamidinium bromide (FABr) with a p-type self-assembled monolayer (SAM) to target the nonradiative recombination pathways. Photoluminescence measurement shows considerable suppression of nonradiative recombination rates after treatment with FABr. However, WBPs still remained susceptible to halide segregation for which the addition of 25% p-type SAM was effective to decelerate segregation. It is observed that FABr can act as a passivating agent of the donor impurities, shifting the Fermi-level (Ef) toward the mid-band gap, while p-type SAM could cause an overweight of Ef toward the valence band. Favorable band bending at the interface could prevent the funneling of carriers toward I-rich clusters. Instead, charge carriers funnel toward an integrated SAM, preventing the accumulation of polaron-induced strain on the lattice. Consequently, n-i-p structured devices with an optimal MFP treatment show an average open-circuit voltage (VOC) increase of about 20 mV and fill factor (FF) increase by 4% compared with the control samples. The unencapsulated devices retained 95% of their initial performance when stored at room temperature under 40% relative humidity for 2800 h.

Thermal Disorder-Induced Strain and Carrier Localization Activate Reverse Halide Segregation.

The reversal of halide ions is studied under various conditions. However, the underlying mechanism of heat-induced reversal remains unclear. This work finds that dynamic disorder-induced localization of self-trapped polarons and thermal disorder-induced strain (TDIS) can be co-acting drivers of reverse segregation. Localization of polarons results in an order of magnitude decrease in excess carrier density (polaron population), causing a reduced impact of the light-induced strain (LIS - responsible for segregation) on the perovskite framework. Meanwhile, exposing the lattice to TDIS exceeding the LIS can eliminate the photoexcitation-induced strain gradient, as thermal fluctuations of the lattice can mask the LIS strain. Under continuous 0.1 W cm⁻2 illumination (upon segregation), the strain disorder is estimated to be 0.14%, while at 80 °C under dark conditions, the strain is 0.23%. However, in situ heating of the segregated film to 80 °C under continuous illumination (upon reversal) increases the total strain disorder to 0.25%, where TDIS is likely to have a dominant contribution. Therefore, the contribution of entropy to the system's free energy is likely to dominate, respectively. Various temperature-dependent in situ measurements and simulations further support the results. These findings highlight the importance of strain homogenization for designing stable perovskites under real-world operating conditions.

Extranodal dissemination of non-Hodgkin lymphoma requires CD47 and is inhibited by anti-CD47 antibody therapy.

Non-Hodgkin lymphoma (NHL) presents as both localized and disseminated disease with spread to secondary sites carrying a worse prognosis. Although pathways driving NHL dissemination have been identified, there are few therapies capable of inhibiting them. Here, we report a novel role for the immunomodulatory protein CD47 in NHL dissemination, and we demonstrate that therapeutic targeting of CD47 can prevent such spread. We developed 2 in vivo lymphoma metastasis models using Raji cells, a human NHL cell line, and primary cells from a lymphoma patient. CD47 expression was required for Raji cell dissemination to the liver in mouse xenotransplants. Targeting of CD47 with a blocking antibody inhibited Raji cell dissemination to major organs, including the central nervous system, and inhibited hematogenous dissemination of primary lymphoma cells. We hypothesized that anti-CD47 antibody-mediated elimination of circulating tumor cells occurred through phagocytosis, a previously described mechanism for blocking anti-CD47 antibodies. As predicted, inhibition of dissemination by anti-CD47 antibodies was dependent on blockade of phagocyte SIRPα and required macrophage effector cells. These results demonstrate that CD47 is required for NHL dissemination, which can be therapeutically targeted with a blocking anti-CD47 antibody. Ultimately, these findings are potentially applicable to the dissemination and metastasis of other solid tumors.

Anterior Skull Base Osteoradionecrosis in the Age of Intensity-Modulated Radiation Therapy: A Case Series.

Background Despites advances in radiation technology, skull base osteoradionecrosis (ORN) continues to be a rare, devastating, and hard to treat complication of radiotherapy. We present three cases of anterior skull base ORN in a cohort treated with intensity-modulated radiation therapy (IMRT). Case Series Three patients developed anterior skull base ORN after receiving at least one round of IMRT. ORN was diagnosed through either nasal endoscopy or imaging findings. The first was a 59-year-old woman with a sinonasal squamous cell carcinoma. Her chemoradiation history was notable for reirradiation and a high dose of radiation (143.3 Gy). The second was a 55-year-old man with recurrent nasopharyngeal carcinoma, whose history was notable for a high dose of radiation (∼140 Gy) and for being reirradiated. The final patient was a 37-year-old woman with an unremarkable history who received radiotherapy (65.0 Gy) for an esthesioneuroblastoma. One patient was asymptomatic and did not receive ORN-specific therapy. The other two were treated with a combination of medical and surgical intervention with successful short-term outcomes (no evidence of infection). Conclusion Anterior skull base ORN can be treated through conservative and surgical means to achieve successful short-term outcomes. Further investigation of long-term outcomes is warranted.

ASSET: Auto-Segmentation of the Seventeen SEgments for Ventricular Tachycardia Ablation in Radiation Therapy.

There has been a recent effort to treat high-risk ventricular tachycardia (VT) patients through radio-ablation. However, manual segmentation of the VT target is complex and time-consuming. This work introduces ASSET, or Auto-segmentation of the Seventeen SEgments for Tachycardia ablation, to aid in radiation therapy (RT) planning. ASSET was retrospectively applied to CTs for 26 thoracic RT patients (13 undergoing VT ablation). The physician-defined parasternal long-axis of the left ventricle (LV) and the axes generated from principal component analysis (PCA) were compared using mean distance to agreement (MDA) and angle of separation. The manually selected right ventricle insertion point and LVs were used to apply the ASSET model to automatically generate the 17 segments of the LV myocardium (LVM). Physician-defined parasternal long-axis differed from PCA by 1.2 ± 0.3 mm MDA and 6.9 ± 0.7 degrees. Segments differed by 0.69 ± 0.29 mm MDA and 0.89 ± 0.03 Dice similarity coefficient. Running ASSET takes <5 min where manual segmentation took >2 h/patient. Agreement between ASSET and expert contours was comparable to inter-observer variability. Qualitative scoring conducted by three experts revealed automatically generated segmentations were clinically useable as-is. ASSET offers efficient and reliable automatic segmentations for the 17 segments of the LVM for target generation in RT planning.

A survival prediction model via interpretable machine learning for patients with oropharyngeal cancer following radiotherapy.

PURPOSE: To explore interpretable machine learning (ML) methods, with the hope of adding more prognosis value, for predicting survival for patients with Oropharyngeal-Cancer (OPC). METHODS: A cohort of 427 OPC patients (Training 341, Test 86) from TCIA database was analyzed. Radiomic features of gross-tumor-volume (GTV) extracted from planning CT using Pyradiomics, and HPV p16 status, etc. patient characteristics were considered as potential predictors. A multi-level dimension reduction algorithm consisting of Least-Absolute-Selection-Operator (Lasso) and Sequential-Floating-Backward-Selection (SFBS) was proposed to effectively remove redundant/irrelevant features. The interpretable model was constructed by quantifying the contribution of each feature to the Extreme-Gradient-Boosting (XGBoost) decision by Shapley-Additive-exPlanations (SHAP) algorithm. RESULTS: The Lasso-SFBS algorithm proposed in this study finally selected 14 features, and our prediction model achieved an area-under-ROC-curve (AUC) of 0.85 on the test dataset based on this feature set. The ranking of the contribution values calculated by SHAP shows that the top predictors that were most correlated with survival were ECOG performance status, wavelet-LLH_firstorder_Mean, chemotherapy, wavelet-LHL_glcm_InverseVariance, tumor size. Those patients who had chemotherapy, with positive HPV p16 status, and lower ECOG performance status, tended to have higher SHAP scores and longer survival; who had an older age at diagnosis, heavy drinking and smoking pack year history, tended to lower SHAP scores and shorter survival. CONCLUSION: We demonstrated predictive values of combined patient characteristics and imaging features for the overall survival of OPC patients. The multi-level dimension reduction algorithm can reliably identify the most plausible predictors that are mostly associated with overall survival. The interpretable patient-specific survival prediction model, capturing correlations of each predictor and clinical outcome, was developed to facilitate clinical decision-making for personalized treatment.