Radiomics;Contemporary Applications in the Management of Anal Cancer;A Systematic Review

Abstract

Introduction

The management of anal cancer relies on clinical and histopathological features for treatment decisions. In recent years, the field of radiomics, which involves the extraction and analysis of quantitative imaging features, has shown promise in improving management of pelvic cancers. The aim of this study was to evaluate the current application of radiomics in the management of anal cancer.

Methods

A systematic search was conducted in Medline, EMBASE, and Web of Science databases. Inclusion criteria encompassed randomized and non-randomized trials investigating the use of radiomics to predict post-operative recurrence in anal cancer. Study quality was assessed using the QUADAS-2 and Radiomics Quality Score tools.

Results

The systematic review identified a total of nine studies, with 589 patients examined. There were three main outcomes assessed in included studies: recurrence (6 studies), progression-free survival (2 studies), and prediction of human papillomavirus (HPV) status (1 study). Radiomics-based risk stratification models were found to provide valuable insights into treatment response and patient outcomes, with all developed signatures demonstrating at least modest accuracy (range: .68-1.0) in predicting their primary outcome.

Conclusion

Radiomics has emerged as a promising tool in the management of anal cancer. It offers the potential for improved risk stratification, treatment planning, and response assessment, thereby guiding personalized therapeutic approaches.

Keywords

radiomics oncology anal cancer survival recurrence treatment response

Introduction

Anal squamous cell carcinoma (ASCC) is relatively rare, accounting for approximately 2.6% of all gastrointestinal cancers.^1,2 However, incidence of ASCC is increasing, particularly in high-income countries.³ ASCC generally has a slow progression and is often responsive to loco-regional treatments due to its low rate of distant metastases at the time of diagnosis.⁴ In recent decades, there has been a notable evolution in the standard of care for non-metastatic disease. The approach shifted from radical surgery to curative treatment with chemo-radiotherapy (CRT). This change in treatment strategy was based on the results of phase III trials conducted in the 1990s.^5,6

Radiomics refers to the conversion of medical imaging into high-throughput, quantifiable data in order to analyze disease patterns, aid decision-making, and predict prognosis.^7,8 Since the concept was introduced in 2012, data from different imaging modalities, including computerized tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET) scans, has been extracted and analyzed.⁹ The goal is to determine patient prognosis and predict outcomes, particularly in the field of surgical oncology.¹⁰ Applying radiomic techniques to predict CRT response in anal cancer patients enables personalized treatment, optimizing outcomes by tailoring therapy to individual characteristics.¹¹ Radiomic nomograms help to stratify patients, identifying those likely to respond well.¹² Predicting recurrence, survival, and disease progression aids treatment decisions, sparing high-risk patients from potentially morbid interventions. An up-to-date literature analysis on radiomics' contemporary use in anal cancer management ensures informed research and practice, harnessing the field's evolving advancements for improved patient care.

The aim of this study is to highlight the current evidence surrounding the contemporary use of radiomics in anal cancer, highlighting strengths, weaknesses, and heterogeneity between studies to guide future research within this novel field of diagnostic and predictive medicine.

Methods

Study Design and Reporting Guidelines

This study is a systematic review of non-randomized trials and follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting guidelines.¹³

Systematic Review Registration

Our systematic review was registered on PROSPERO in March 2023 (ID: CRD42023456092).

Search Strategy

The following databases were searched as part of the systematic review in July 2023: EMBASE, Medline, and Web of Science. The last date of search was 18^th July 2023. The grey literature was also searched to further identify other suitable publications. The systematic search process with detailed search terms are outlined in the supplementary material S1.

Inclusion/exclusion Criteria

Inclusion Criteria

- Patient aged 18 years old and above

- Studies investigating the use of radiomics to predict oncological outcomes in patients with anal cancer

- Prospective or Retrospective Studies.

Exclusion criteria

- Case series/reports.

- Consensus statements.

- Conference abstracts.

- Missing/conflicting data with no response from contacted authors.

Study Selection, Data Extraction and Critical Appraisal

A database was created using the reference managing software EndNote X9^TM. Two researchers (HCT and NOS) reviewed outputs from the searches independently of each other.

Initially, duplicates were removed. Study titles were then screened and assessed for potential relevance. The abstracts of selected potential studies were then read and assessed for eligibility for inclusion, based on the inclusion/exclusion criteria detailed above. Rejected studies were grouped together in the database by their reason for exclusion. The full texts of the abstracts deemed eligible for inclusion were then further analyzed using the same criteria.

In order to extract and store data efficiently, the Cochrane Collaboration screening and data extraction tool, Covidence, was used.¹⁴ Data was collected by two reviewers (HCT and NOS) independently, using the following headings; study details, study design, population, intervention, comparison groups, and outcomes.

A critical appraisal of the methodological quality and risk of bias of the included studies was performed. The critical appraisal was completed by two reviewers independently. Quality assessment of the included studies was performed according the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) and Radiomics Quality Score (RQS).^10,15

Results

Search Results

The literature search yielded a total of 900 results. Following the removal of 450 duplicates, 450 studies were screened. After the initial screen, 237 abstracts were reviewed and assessed for eligibility, of which twelve were selected for full text review. From these twelve full texts, a total of nine studies met the inclusion criteria and were included in our analysis.^16-24 See Figure 1.

Figure 1.

A PRISMA Flowchart of the selection of relevant publications included in this review.

Methodological Characteristics, Quality of Studies, Patient Characteristics

Overall, six of the included studies were retrospective in nature,^{16-18,21,22,24} with the remaining three being prospective.^19,20,23 In total, four studies were multicentre,^17,20,21,25 with the other five being single-center studies.^{16,18,19,22,23} The total number of participants from the nine included studies was 589.

Data quality, assessed using the RQS and QUADAS-2 tools, was generally satisfactory. All studies were deemed low risk of bias based off the QUADAS-2 tool, while 66.6% (6/9) of included studies received an RQS score >30%. A detailed explanation of the tools and breakdown of the results can be found in supplementary material (S2-S5). Table 1 summarizes the methodological characteristics of the included studies.

Table 1.

Methodological Characteristics of the Included Studies.

Study	Country	Journal	Study Design	Multicentre	Primary Outcome
Hocquelet et al (2018)	France	European Radiology	Retrospective	No	Progression free survival
Owczarczyk et al (2019)	United Kingdom	Radiotherapy and Oncology	Retrospective	No	Predict recurrence
Rusten et al (2019)	Norway	British Journal of Radiology	Prospective	No	Predict loco-regional recurrence
Jones et al (2019)	Australia	Radiation Oncology	Prospective	No	Predict local and any recurrence
Brown et al (2019)	UK	European Journal of Nuclear Medicine and Molecular Imaging	Retrospective	No	Progression free survival
Wang et al (2019)	United States	Frontiers in Oncology	Retrospective	Yes	Predict recurrence
Jones et al (2020)	Australia	Radiotherapy and Oncology	Prospective	Yes	Predict local and distant recurrence
Giraud et al (2021)	France	Cancers	Retrospective	Yes	Predict recurrence
Leijenaar et al (2023)	Belgium	Nature Scientific Reports	Retrospective	Yes	Predict p16 (human papillomavirus) of anal cancer patients

Acquisition Parameters

PET-CT was the most commonly used modality, implemented in six studies overall.^{16,18,19,22-24} The remaining studies developed their models from T2-weighted MRI (2/9) or contrast-enhanced CT (1/9).^17,20,21 Acquisition parameters are illustrated in Table 2.

Table 2.

Summary of Imaging Parameters and Radiomics for Prediction of Response to Chemo-Radiotherapy in Anal Squamous Cell Carcinomas.

Authors Main Results	Number of Patients	Imaging Modality	Segmentation	Selected Imaging Parameters	Prediction Algorithm	Conclusions
Hocquelet et al (2018)	28	Pre-CRT 1.5T MRI (axial T2w)	Manual segmentation	First and second-order (GLCM) RCS	Cox regression	MRI texture analysis provides prognostic factors of event occurrence
Owczarczyk et al (2019)	40	Pre and post-CRT 1.5 T MRI (T2w, DW)	Manual segmentation	Tumor volume, first, second-order (GLCM) and fractal RCS	Risk factors for variables selection logistic regression	MRI heterogeneity analysis offers complementary information, in addition to clinical staging, in predicting outcome
Rusten et al (2019)	93	Pre-CRT PET-CT (+/- at W2 of CRT for 39 patients)	Manual segmentation	SUV (max, mean, peak), MTV, TLG, ZMP	Cox regression	Pre-treatment PET parameters are predictive of chemoradiotherapy outcome in anal cancer, although HPV negativity and N3 stage are the strongest single predictors. Predictions can be improved by combining HPV with PET parameters such as MTV, TLG or ZMP.
Jones et al (2019)	19	Pre-CRT and W12 post-CRT PET-CT	Manual segmentation	SUV (max, mean, median, standard deviation and peak), MTV, TLG	Logistic regression	FDG-PET parameters are predictive of recurrence in AC.
Brown et al (2019)	189	Pre-CRT PET-CT	Semi-automatic segmentation	First, second and third-order RCS	Logistic regression	Good performance in prediction of PFS
Wang et al (2019)	17	PET/CT at baseline (Pre-CRT), in the middle of the treatment (Mid-CRT) and post-treatment (Post-CRT)	Manual segmentation	Energy, entropy, correlation, inverse difference moment, inertia, cluster shade, cluster prominence, Haralick correlation.	Logistic regression	Experimental results demonstrated the potential of serial PET/CT scans in early prediction of anal tumor recurrence
Jones et al (2020)	25	Pre-CRT, W2, W4 of CRT and W8 post-CRT 3T MRI (DWI, DCE)	Manual segmentation	First-order RCS	Logistic regression	Several ADC parameters at various time points correlate with recurrence suggesting DW-MRI is a potential biomarker for ASCC. (Baseline skewness, week 2 skewness, week 4 kurtosis)
Giraud et al (2021)	82	Pre-CRT CT	N/a	FirstOrder_Entropy and GLCM_JointEnergy	Logistic regression	Mixed model with two clinical and two radiomic features was predictive of 2-year disease control after CRT and could contribute to identify high-risk patients amenable to treatment intensification
Leijenaar et al (20230	96	Pre-CRT MRI axial T2-weighted (T2w)	Manual segmentation	N/a	Logistic regression	Reasonable performance at predicting HPV status in anal cancer

CRT: chemoradiotherapy, GLCM: gray-level co-occurrence matrices, RCS: radar cross section, MRI; magnetic resonance imaging, CT: Computerized tomography, PET: positron emission tomography, SUV: standardized uptake value, MTV: Metabolic tumor volume, TLG: total lesion glycolysis, ZMP: Z-normalized combination of MTV and SUVpeak, DW: Diffusion-weighted, PFS: progression free survival, DCE: dynamic contrast-enhanced, ASCC: anal squamous cell carcinoma, ADC: apparent diffusion coefficient, N/a: not applicable.

Development of Signatures

Exact feature extraction methods varied across studies; however, a relatively similar pathway was followed across the board. Regions of interest (ROI) were first segmented manually by experienced radiologists in all included studies. Radiomic features were then extracted from these segments using radiomics software. Specific software used for image segmentation and feature extraction is demonstrated in Table 2. Logistic and Cox regression models were subsequently applied to features to remove sub-optimal features. Remaining features were used to construct individual radiomics signatures.

Primary Outcomes

There were three main outcomes assessed in included studies (see table 1):

(i) Recurrence (6/9 studies)^19,20,22-24

(ii) Progression-free survival (2/9 studies)^16,18

(iii) Prediction of human papillomavirus (HPV) status (1/9 studies)²¹

Disease recurrence

Giraud et al¹⁷ aimed to identify novel prognostic factors that could effectively predict cancer recurrence within a 2-year timeframe after CRT. A pre-therapeutic MRI radiomic analysis was utilized, alongside conventional clinical variables, to construct a logistic regression model capable of forecasting the likelihood of recurrence. The study population comprised 82 patients, randomly divided into training (n = 54) and testing (n = 28) subsets. Following the selection of optimal variables, the logistic regression model incorporates two radiomic features (FirstOrder_Entropy and GLCM_JointEnergy) and two clinical features (tumor size and CRT length). Among the 82 patients included, 24 (29%) experienced disease relapse at the 2-year follow-up. Univariate analysis of the training set showed that two clinical (tumor size and CRT length) and two radiomic features (FirstOrder_Entropy and GLCM_JointEnergy) were associated with disease control and were therefore included in the model. A mixed model incorporating both clinical and radiomic features outperformed the clinical model in both the training (area under curve (AUC) .825 vs .758, accuracy 87% vs 75.9%) and testing (AUC .898 vs .714, accuracy 85.7% vs 78.6%) sets, allowing stratification of patients into high and low-risk groups for disease relapse (HR 8.60, P = .005). The combined clinical and radiomic model demonstrated predictive potential for 2-year disease control following CRT in ASCC patients.

Owczarcyk et al assessed the potential of image heterogeneity analysis of standard care MRI in predicting the outcomes of CRT in patients diagnosed with ASCC.²² The researchers outlined the whole tumor 3D volume-of-interest (VOI) on pre- and post-treatment T2-weighted (T2w) and diffusion-weighted imaging (DWI) scans. They then identified independent imaging features that showed the highest predictive potential for disease recurrence. These imaging features were combined with the baseline clinico-pathological model. To assess the value of the extended models, the researchers used the net reclassification improvement (NRI) algorithm. The study included 40 patients who underwent 1.5 T pelvic MRI at baseline as well as post-CRT. The combination of two baseline MR heterogeneity features (baseline T2w energy and DWI coefficient of variation) demonstrated the highest predictive capability for disease recurrence, leading to a significant net reclassification index (P < .001). Cross-validation analysis further confirmed the effectiveness of these imaging variables, resulting in a 34.8% reduction in percentage error for the primary endpoint and an 18.1% reduction for the secondary endpoint when added to the baseline model.

Wang et al²⁴ conducted PET/CT scans at three different time points: baseline (Pre-CRT), during treatment (Mid-CRT), and post-treatment (Post-CRT). Image features were extracted from the manually delineated gross tumor volume, encompassing geometry, intensity, and texture features. The differences in feature values between two time points were also computed and analyzed. Various models, such as univariate logistic regression, multivariate models, and naïve Bayesian classifiers, were employed to identify significant predictors of tumor recurrence. The predictive accuracy was evaluated using the area under the receiver operating characteristic (ROC) curve (AUC). The study found specific geometry, intensity, and texture features that were associated with tumor recurrence in the univariate analysis. Moreover, certain features, such as “Roundness” in Post-CRT and Pre-CRT CT scans, demonstrated predictive capabilities with an AUC value of 1.00 in the multivariate logistic regression model. Additionally, early prediction ability was investigated, and combinations of features from Pre-CRT and Mid-CRT scans were identified as valuable predictors. For instance, the combination of “Orientation” in Pre-CRT standard uptake value (SUV), “Mean” in Pre-CRT CT, and “Mean of Long Run High Gray Level Emphasis” in Pre-CRT CT yielded an AUC of 1.00.

Jones et al prospectively investigated 19 patients with non-metastatic anal cancer between 2013 and 2017.¹⁹ Each patient underwent FDG-PET/CT imaging both before CRT and 12 weeks after CRT. The researchers generated a volume of interest (VOI) around the primary tumor using six different SUV thresholds and extracted various PET parameters, including SUV max, mean, median, standard deviation, peak, metabolic tumor volume (MTV), and total lesion glycolysis. The predictive value of these parameters was analyzed using exact logistic regression and ROC AUC analyses at each time point. The results showed that certain FDG-PET parameters, particularly median SUV on the post-CRT PET scan, were significantly correlated with local recurrence (P < .01) and exhibited excellent discriminatory power (ROC AUC 1.00). Moreover, the MTV bounded by a threshold of 41% SUVmax on the pre-CRT PET scan was predictive of any recurrence (P = .03) and showed excellent discriminatory ability (ROC AUC .89).

Rusten et al²³ evaluated the predictive role FDG PET in assessing locoregional failure among 93 ASCC patients undergoing CRT. Before treatment, all patients underwent FDG-PET/CT scans, with a subgroup of 39 patients receiving additional scans two weeks into the treatment. The analysis focused on various FDG-PET parameters, such as standardized uptake values (SUVmax/peak/mean), metabolic tumor volume (MTV), total lesion glycolysis (TLG), and a Z-normalized combination of MTV and SUVpeak (ZMP). Univariate and bivariate Cox regression analysis were employed, alongside clinicopathological factors like tumor and lymph node stage, gross tumor volume (GTV), and human papillomavirus (HPV) status, to predict locoregional failure. The findings revealed that N3 lymph node stage, HPV-negative tumor status, GTV, MTV, TLG, and ZMP were significant predictors of locoregional failure (P < .01), while SUVmax/peak/mean did not exhibit significant predictive value (P > .2). In combination with N3 stage, ZMP, TLG, and MTV, HPV status emerged as the most independent predictor in bivariate analysis (P < .02). Moreover, FDG-PET parameters measured two weeks into radiotherapy showed a decrease of 30%-40% compared to the initial values; however, neither the absolute nor relative decrease in these parameters improved the prediction models for locoregional failure.

Jones et al²⁰ investigated 25 patients with non-metastatic ASCC in a multi-center prospective study. MP-MRIs, incorporating diffusion-weighted MRI (DW-MRI) and dynamic contrast-enhanced MRI (DCE-MRI) sequences, were conducted at baseline, during the second and fourth weeks of CRT, and 8 weeks post-treatment completion. Various histogram metrics were derived from the MP-MRI data, including maximum, mean, median, minimum, skewness, kurtosis, and standard deviation. The analysis revealed strong associations between certain apparent diffusion coefficient (ADC) metrics obtained from DW-MRI and local recurrence of ASCC, displaying excellent discriminatory capability. Notably, baseline skewness and standard deviation, as well as week 2 skewness and standard deviation, and week 4 kurtosis and standard deviation, were identified as prognostic indicators for local recurrence. Furthermore, changes in minimum ADC between baseline and week 2, and between baseline and week 4, were predictive of local recurrence. Regarding any recurrence, significant predictive factors included ADC minimum and standard deviation at baseline, as well as ADC maximum and standard deviation at week 4.

Progression-Free Survival

Brown et al¹⁶ assessed novel imaging features extracted from pre-treatment FDG-PET/CT scans for predicting progression free survival (PFS) in ASCC patients. Radiomic feature extraction was performed using LIFEx software on baseline FDG-PET/CT images. Logistic regression models were generated using elastic net regularization and feature selection on a randomly selected training cohort, and the models were then applied to a validation cohort. Performance comparison of the models was conducted using ROC-AUC analysis, including a model with standard clinical prognostic factors (age, sex, tumor, and nodal stage—Model A), a radiomic feature model (Model B), and a combined radiomic/clinical model (Model C). The study comprised 189 patients, with 145 in the training cohort and 44 in the validation cohort. Median follow-up periods were 35.1 and 37.9 months, respectively, with 70.3% and 68.2% reaching the specified follow-up time with PFS. The prognostic model included the following radiomic features: GLCM entropy (a measure of randomness in the distribution of co-occurring pixel grey-levels), NGLDM busyness (a measure of spatial frequency of intensity changes between nearby voxels with different grey levels), minimum CT value (lowest Hounsfield Unit within the lesion), and SMTV (a standardized version of metabolic tumor volume), alongside tumor and nodal stage. The AUCs for Model A (clinical), Model B (radiomic), and Model C (radiomic/clinical) were .6355, .7403, and .7412 in the training cohort, and .6024, .6595, and .7381 in the validation cohort, respectively.

Hocquelet et al¹⁸ aimed to investigate whether regular MRI findings and tumor texture features on pre-CRT imaging could serve as potential predictors of event-free survival (defined as disease progression or death) in patients with anal squamous cell carcinoma (ASCC) without metastasis. The study included 28 patients who underwent CRT for pathologically confirmed ASCC and had pre-CRT MRI images. Texture analysis was performed on axial T2-weighted images, using a 3D region of interest delineated around the entire tumor volume. First-order analysis quantified histogram-based features, and second-order statistical texture features were derived from the grey-level co-occurrence matrix at distances of 1 (d1), 2 (d2), and 5 (d5) pixels. Prognostic factors were evaluated using Cox regression, and model performance was assessed using the Harrell C-index.

Among the 28 patients, eight experienced tumor progression, leading to six tumor-specific deaths. After adjusting for age, gender, and tumor grade, two texture features showed associations with event occurrence: skewness (HR = .131, 95% CI = 0-.447, P = .005) and cluster shade_d1 (HR = .601, 95% CI = 0-.861, P = .027). The corresponding Harrell C-indices for these associations were .846 (95% CI = .697-.993) and .851 (95% CI = .708-.994), respectively.

Prediction of p16 (HPV) Status in Anal Cancer

Prediction of p16 (HPV) status was assessed by Leijenaar et al²¹ Their paper focused on evaluating the performance of an existing CT-based radiomic signature, originally designed for oropharyngeal cancer, to predict HPV status in the context of anal cancer. Their dataset was made up of 59 patients from two different medical centers. The primary objective was to assess whether the radiomic signature could effectively predict HPV status based on p16 immunohistochemistry. After conducting predefined statistical tests, the researchers obtained an AUC value of .68 [95% CI (.32-1.00)] and an F1 score of .78 for the anal cancer dataset. The radiomic signature achieved a TRIPOD level 4 (57%) with an RQS of 61%. The study successfully demonstrates that this radiomic signature has the potential to identify the clinically relevant molecular phenotype, that is, HPV-ness, across multiple cancers, thereby serving as a promising CT imaging biomarker for p16 status.

Discussion

Factors such as male gender, N-positive stage, and tumor length > 5 cm are prognostic indicators in ASCC, correlating with less favorable clinical outcomes.^25,26 Additional potential prognostic markers include: neutrophil-to-lymphocyte ratio for locoregional recurrence, age >55, skin ulceration, metastatic inguinal nodes, and a total RT dose >60 Gy.^26-28 The creation of predictive nomograms has yielded promising outcomes in forecasting cancer-specific and overall survival rates, as well as aiding risk assessment; however, it is crucial to acknowledge that these nomograms have been exclusively grounded in clinical data.^29-31 Notably, the promise of radiomics features extracted from medical imaging, introducing the potential for supplementary insightful data. The amalgamation of these novel parameters with pre-existing clinical factors appears to augment performance.^16,17,22,23 This is precisely where the prowess of machine learning (ML) algorithms becomes evident, having already demonstrated remarkable accomplishments across various cancer categories.^32-35

A proportion of ASCC patients undergoing curative CRT still require salvage abdominoperineal resection (APR) due to inadequate treatment response or local recurrence.^36,37 In this context, the integration of radiomics and ML from outset imaging may play a significant role in stratifying patients who are at higher risk of requiring salvage APR. With radiomics and ML the potential to identify early indicators of treatment resistance or local recurrence, facilitating timely interventions and personalized treatment strategies.^19,20,22,23 Consequently, such predictive models may not only enhance clinical decision-making, improving patient outcomes and refining therapeutic approaches in the management of anal cancer.

This review establishes the feasibility and efficacy of formulating radiomic signatures for the prediction of oncological outcomes in patients with anal cancer, based on pre/post-operative imaging data. Inclusion of various studies revealed that all developed signatures demonstrated at least modest accuracy (range: .68-1.0) in predicting their primary outcome, with one study reporting an exceptionally high AUC of 1.0.²⁴ However, these encouraging results are constrained by several limitations, predominantly arising from the substantial variability in imaging protocols and radiomic feature extraction methodologies.

We highlight that the use of radiomics (and radiogenomics) in the field of ASCC is still novel, but the potential is massive given the natural progression of this disease. Further refinement of radiomic methodologies will align assessment and reporting of predictive radiomics signatures.

Conclusion

The integration of radiomics into the management of anal cancer represents a potential adjunct into management, highlighting those whom have higher risk of treatment failure requiring salvage surgery. Standardized protocols, larger multicentre studies, and prospective validation are essential to underscore its role in clinical decision-making.

Supplemental Material

Supplemental Material - Radiomics; Contemporary Applications in the Management of Anal Cancer; A Systematic Review

Supplemental Material for Radiomics; Contemporary Applications in the Management of Anal Cancer; A Systematic Review by Hugo C. Temperley, Niall J. O’Sullivan, Caitlin Waters, Alison Corr, Brian J. Mehigan, Grainne O’Kane, Paul McCormick, Charles Gillham, Emanuele Rausa, John O. Larkin, James F. Meaney, Ian Brennan, and Michael E. Kelly in The American Surgeon.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Supported by Joly Cancer Leadership Programme, Trinity St James Cancer Institute.

ORCID iD

Hugo C. Temperley

Supplemental Material

Supplemental material for this article is available online.

References

Pessia

Romano

Giuliani

Lazzarin

Carlei

Schietroma

. Squamous cell anal cancer: management and therapeutic options. Ann Med Surg (Lond). 2020;55:36-46. doi:10.1016/j.amsu.2020.04.016

Siegel

Miller

Jemal

. Cancer statistics, 2020. CA Cancer J Clin. 2020;70(1):7-30. doi:10.3322/caac.21590

Islami

Ferlay

Lortet-Tieulent

Bray

Jemal

. International trends in anal cancer incidence rates. Int J Epidemiol. 2017;46(Issue 3):924-938. doi:10.1093/ije/dyw276

Carr

Jin

Hubbard

. Research on anal squamous cell carcinoma: systemic therapy strategies for Anal Cancer. Cancers (Basel). 2021;13(9):2180. doi:10.3390/cancers13092180

Bosset

Pavy

Roelofsen

Bartelink

. Combined radiotherapy and chemotherapy for anal cancer EORTC radiotherapy and gastrointestinal cooperative groups. Lancet. 1997;349:205-206.

UKCCCR . Epidermoid anal cancer: results from the UKCCCR randomised trial of radiotherapy alone versus radiotherapy, 5-fluorouracil, and mitomycin. UKCCCR Anal Cancer Trial Working Party. UK Co-ordinating Committee on Cancer Research. Lancet. 1996;348:1049-1054.

Guiot

Vaidyanathan

Deprez

, et al. A review in radiomics: making personalized medicine a reality via routine imaging. Med Res Rev. 2022;42:426-440.

Gillies

Kinahan

Hricak

. Radiomics: images are more than pictures, they are data. Radiology. 2016;278:563-577.

Liu

Wang

Dong

, et al. The applications of radiomics in precision diagnosis and treatment of oncology: opportunities and challenges. Theranostics. 2019;9:1303-1322

10.

Lambin

Leijenaar

RTH

Deist

, et al. Radiomics: the bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol. 2017;14:749-762.

11.

Giraud

Sargos

Leduc

Saut

Vuong

Vendrely

. Radiomics and machine learning in anal squamous cell carcinoma: a new step for personalized medicine? Appl Sci. 2020;10:1988. doi:10.3390/app10061988

12.

Stanzione

Verde

Romeo

Boccadifuoco

Mainenti

Maurea

. Radiomics and machine learning applications in rectal cancer: current update and future perspectives. World J Gastroenterol. 2021;27:5306-5321. doi:10.3748/wjg.v27.i32.5306

13.

Liberati

Altman

Tetzlaff

, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. Ann Intern Med. 2009;151(4):W65-W94. doi:10.7326/0003-4819-151-4-200908180-00136

14.

Covidence systematic review software. Melbourne, Australia: Veritas Health Innovation; 2023

15.

Whiting

Rutjes

Westwood

, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med. 2011;155:529-536.

16.

Brown

Zhong

Frood

, et al. Prediction of outcome in anal squamous cell carcinoma using radiomic feature analysis of pre-treatment FDG PET-CT. Eur J Nucl Med Mol Imaging. 2019;46(13):2790-2799. doi:10.1007/s00259-019-04495-1

17.

Giraud

Saut

Aparicio

, et al. MRI-based radiomics input for prediction of 2-year disease recurrence in anal squamous cell carcinoma. Cancers (Basel). 2021;13(2):193. doi:10.3390/cancers13020193

18.

Hocquelet

Auriac

Perier

, et al. Pre-treatment magnetic resonance-based texture features as potential imaging biomarkers for predicting event free survival in anal cancer treated by chemoradiotherapy. Eur Radiol. 2018;28(7):2801-2811. doi:10.1007/s00330-017-5284-z

19.

Jones

Hruby

Metser

, et al. FDG-PET parameters predict for recurrence in anal cancer - results from a prospective, multicentre clinical trial. Radiat Oncol. 2019;14(1):140. doi:10.1186/s13014-019-1342-9

20.

Jones

Hruby

Coolens

, et al. A prospective, multi-centre trial of multi-parametric MRI as a biomarker in anal carcinoma. Radiother Oncol. 2020;144:7-12. doi:10.1016/j.radonc.2019.10.001

21.

Leijenaar

RTH

Walsh

Aliboni

, et al. External validation of a radiomic signature to predict p16 (HPV) status from standard CT images of anal cancer patients. Sci Rep. 2023;13(1):7198. doi:10.1038/s41598-023-34162-3

22.

Owczarczyk

Prezzi

Cascino

, et al. MRI heterogeneity analysis for prediction of recurrence and disease free survival in anal cancer. Radiother Oncol. 2019;134:119-126. doi:10.1016/j.radonc.2019.01.022

23.

Rusten

Rekstad

Undseth

, et al. Anal cancer chemoradiotherapy outcome prediction using 18F-fluorodeoxyglucose positron emission tomography and clinicopathological factors. Br J Radiol. 2019;92

24.

Wang

Zhang

Chuong

, et al. Prediction of anal cancer recurrence after chemoradiotherapy using quantitative image features extracted from serial 18F-FDG PET/CT. Front Oncol. 2019;9:934. doi:10.3389/fonc.2019.00934

25.

Myerson

Kong

Birnbaum

, et al. Radiation therapy for epidermoid carcinoma of the anal canal, clinical and treatment factors associated with outcome. Radiother Oncol. 2001;61:15-22.

26.

Faivre

Peiffert

Vendrely

, et al. Prognostic factors of colostomy free survival in patients presenting with locally advanced anal canal carcinoma: a pooled analysis of two prospective trials (KANAL 2 and ACCORD 03). Radiother Oncol. 2018;129(3):463-470. doi:10.1016/jradonc201808008

27.

Glynne-Jones

Saleem

Harrison

Mawdsley

Hall

. Background and current treatment of squamous cell carcinoma of the anus the Oncol Ther. 2016;4(2):135-172 doi:10.1007/s40487-016-0024-0

28.

Glynne-Jones

Sebag-Montefiore

Adams

, et al. Prognostic factors for recurrence and survival in anal cancer: generating hypotheses from the mature outcomes of the first United Kingdom Coordinating Committee on Cancer Research Anal Cancer Trial (ACT I). Cancer. 2013;119(4):748-755. doi:10.1002/cncr27825

29.

Han

Zhu

. Survival prediction models for patients with anal carcinoma receiving definitive chemoradiation: a population-based study. Oncol Lett. 2020;19:1443-1451.

30.

Tang

Zhu

Huang

, et al. Development and validation of prognostic survival nomograms for patients with anal canal cancer: a SEER-based study. Int J Gen Med. 2021;14:10065-10081.

31.

Yang

Wang

Chen

Zhou

. Construction and validation of a prognostic nomogram for anal squamous cell carcinoma. Cancer Med. 2022;11(2):392-405. doi:10.1002/cam44458

32.

Afrash

Shafiee

Kazemi-Arpanahi

. Establishing machine learning models to predict the early risk of gastric cancer based on lifestyle factors. BMC Gastroenterol. 2023;23(1):6. doi:10.1186/s12876-022-02626-x

33.

Burnett

Zhou

Brophy

, et al. Machine learning in colorectal cancer risk prediction from routinely collected data: a review. Diagnostics (Basel). 2023;13(2):301. doi:10.3390/diagnostics13020301

34.

Yang

Jiang

Luo

. Machine learning for lung cancer diagnosis, treatment, and prognosis. Genomics Proteomics Bioinformatics. 2022;20(5):850-866. doi:10.1016/j.gpb.2022.11.003

35.

Rabiei

Ayyoubzadeh

Sohrabei

Esmaeili

Atashi

. Prediction of breast cancer using machine learning approaches. J Biomed Phys Eng. 2022;12(3):297-308. doi:10.31661/jbpe.v0i0.2109-1403

36.

Hagemans

JAW

Blinde

Nuyttens

, et al. Salvage abdominoperineal resection for squamous cell anal cancer: a 30-year single-institution experience. Ann Surg Oncol. 2018;25(7):1970-1979. doi:10.1245/s10434-018-6483-9

37.

Shakir

Adams

Cooper

, et al. Patterns and predictors of relapse following radical chemoradiation therapy delivered using intensity modulated radiation therapy with a simultaneous integrated boost in anal squamous cell carcinoma. Int J Radiat Oncol Biol Phys. 2020;106(2):329-339. doi:10.1016/j.ijrobp.2019.10.016

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.61 MB