(AGENPARL) - Roma, 19 Gennaio 2026 - (AGENPARL) – Mon 19 January 2026 Article
MRI-Based Morphological Features as Predictors of Clinical
Outcomes in Locally Advanced Rectal Cancer Treated with
Neoadjuvant Chemoradiotherapy: Insights from a
Single-Institution Experience
Marco Lucarelli 1,*, Consuelo Rosa 2, Giulia de Pasquale 2, Monica Di Tommaso 2, Tamara Santone 2,
Antonietta Augurio 2, Angelo Di Pilla 2, Marianna Nuzzo 2, Maria Taraborrelli 2, Marianna Trignani 2,
Annamaria Vinciguerra 2, Andrea Delli Pizzi 3,4, Marta Di Nicola 5, Domenico Genovesi 2,5,† and
Andrea D’Aviero 2,5,†
Division of Radiation Oncology, European Institute of Oncology IRCSS, 20141 Milan, Italy
Department of Radiation Oncology, “S.S. Annunziata” Chieti Hospital, 66100 Chieti, Italy;
3 Department of Innovative Technologies in Medicine and Dentistry, “G. d’Annunzio” University,
4 ITAB Institute for Advanced Biomedical Technologies, “G. d’Annunzio” University, 66100 Chieti, Italy
5 Department of Medical, Oral and Biotechnological Sciences, “G. d’Annunzio” University of Chieti,
† These authors contributed equally to this work.
Abstract
Academic Editor: Francesco
Pucciarelli
Received: 15 December 2025
Revised: 31 December 2025
Accepted: 4 January 2026
Published: 6 January 2026
Copyright: © 2026 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
Objectives: This study evaluates MRI-based morphological features as predictors of longterm clinical outcomes in patients with locally advanced rectal cancer (LARC) treated with
neoadjuvant chemoradiotherapy (CRT). Methods: A retrospective analysis was performed on 134 patients treated between 2014 and 2024. Patients underwent dose-intensified radiotherapy (55 Gy) with concurrent capecitabine followed by surgery. Radiological
features analyzed on pre- and post-CRT MRI included Tumor Extension Beyond Muscularis Propria (TEMP), Circumferential Resection Margin (CRM), Extramural Venous Invasion (EMVI), and Lateral Lymph Nodes (LLN). Results: Five-year Overall Survival (OS),
Disease-Free Survival (DFS), and Local Control (LC) rates were 85%, 83%, and 88%, respectively. Patients with TEMP > 5 mm had significantly worse LC (p = 0.02) and DFS (p
= 0.04). A positive CRM ( 5 mm, CRM < 1 mm, EMVI, and pathological LLN are significant predictors of worse oncological outcomes. Identifying these
imaging biomarkers allows for better risk stratification and personalized treatment strategies in LARC.
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license.
J. Clin. Med. 2026, 15, 430
Keywords: rectal cancer; magnetic resonance imaging; outcomes prediction
J. Clin. Med. 2026, 15, 430
2 of 20
1. Introduction
1.1. Background
Globally, colorectal cancer ranks as the third most frequently diagnosed malignancy
and represents the second most common cause of cancer mortality [1]. Throughout Europe in 2020, this disease comprised 12.7% of newly diagnosed cancers and accounted for
12.4% of cancer-related deaths [2]. Within Italy, it represented 11.6% of all new cancer
diagnoses and 10.8% of all cancer deaths during the same timeframe [3].
For patients with locally advanced rectal cancer (LARC), defined as stage II (tumor
invading the muscular wall without lymph node involvement) and III (with lymph node
involvement), specifically cT3 N0 and cT3-4 N0-2 M0 tumors, the established treatment
approach involves long-course preoperative chemoradiotherapy (CRT) followed by surgical resection. The landmark German CAO/ARO/AIO-94 phase III randomized trial
demonstrated that a preoperative approach achieves high local control (LC) rates (5-year
LC of 94%), higher sphincter preservation rates (39% vs. 19%), and lower incidence of both
acute (27% vs. 40%, p = 0.001) and late toxicity (14% vs. 24%) compared to postoperative
treatment [4]. However, no difference was observed between the two treatment schedules
regarding disease-free survival (DFS) (36% vs. 38%) and overall survival (OS) (76% vs.
74%) [4].
Artificial intelligence (AI), encompassing machine learning (ML), artificial neural
networks (ANNs), and deep learning (DL), has emerged as a transformative force in medical imaging and oncology [5]. ML involves training algorithms on data to make predictions without explicit programming, while DL uses multi-layered neural networks to process complex imaging datasets [5]. In rectal cancer imaging, AI applications span two complementary approaches: (1) traditional radiological assessment of morphological features
based on established guidelines, and (2) computational radiomics involving automated
extraction of quantitative imaging features through ML/DL algorithms.
True radiomics, as distinct from conventional morphological assessment, involves
high-throughput extraction of hundreds of quantitative features from medical images—
including texture patterns, shape descriptors, intensity distributions, and higher-order
statistical features—that may not be visually perceptible to human observers [6–11].These
features are then analyzed using ML algorithms to develop predictive models. Recent DL
approaches utilizing convolutional neural networks (CNNs), such as ResNet-101 and Vision Transformers (ViT), have demonstrated superior performance in predicting lymph
node metastasis and other oncological outcomes in rectal cancer, achieving areas under
the curve (AUC) ranging from 0.819 to 0.961, significantly outperforming conventional
radiologist assessments [6,8]. These AI-driven approaches enhance model interpretability
through tools such as Shapley Additive exPlanations (SHAP), allowing clinicians to understand the decision-making process underlying predictions [9]. Studies have demonstrated that radiomic models combining multiparametric MRI features can predict distant
metastasis with C-indices up to 0.775 [8] and enhance the predictive accuracy of OS from
0.672 using only clinical features to 0.730 when incorporating radiomic features [11].
However, clinical implementation of AI-based radiomics faces several challenges, including: (1) the need for large, standardized training datasets, (2) computational expertise
and infrastructure requirements, (3) limited interpretability of black-box models, and (4)
validation across different institutions and MRI protocols [5]. While tools such as SHAP
enhance model interpretability [9], the clinical integration of fully automated radiomic
models remains an evolving field.
In parallel, conventional morphological MRI features—such as those evaluated in
our study—represent established, guideline-supported imaging biomarkers that are routinely assessed in clinical practice [12,13]. These features, while based on visual
J. Clin. Med. 2026, 15, 430
3 of 20
assessment rather than computational extraction, provide clinically validated prognostic
information and inform treatment decisions in real-world settings. The European Society
of Gastrointestinal and Abdominal Radiology (ESGAR) and other international guidelines
mandate reporting of these morphological characteristics for rectal cancer staging [13].
Despite the proven efficacy of neoadjuvant treatment demonstrated since 2012 [4],
clinical practice reveals that patients with the same clinical stage treated with the same
therapeutic protocol have different outcomes. This heterogeneity in treatment response
highlights the need for better predictive biomarkers, whether derived from conventional
morphological assessment or emerging AI-based techniques.
1.2. Tumor Extension Beyond Muscularis Propria (TEMP)
Tumors receive T3 classification when breach of the muscularis propria occurs with
tumor extension beyond the rectal wall into the mesorectal fat. T3 lesions undergo further
subdivision into four categories determined by the distance from the muscularis propria
to maximum extramural invasion: T3a 15
mm. However, tumors spreading ≤ 5 mm into the mesorectum have local recurrence and
survival rates similar to those limited to the muscularis propria (T2) [12]. Indeed, as documented by Siddiqui and colleagues in their meta-analysis, patients with extension beyond the muscularis propria > 5 mm showed lower rates of OS and DFS and LC [14].
1.3. Circumferential Resection Margin (CRM)
Among the most critical imaging markers for predicting local recurrence and serving
as a reliable prognostic factor for DFS and OS is the circumferential resection margin
(CRM), characterized as a distance of 1 mm or less from the tumor or irregular lymph
nodes to the mesorectal fascia, as established by the European Society of Gastrointestinal
and Abdominal Radiology guidelines [13] and the North American Society of Abdominal
Radiology [15]. This marker plays such an important role because the outcome of the
standard total mesorectal excision (TME) surgery depends on the relationship of the tumor to the mesorectal fascia. The MERCURY study, a prospective multicenter study, analyzed 374 patients and observed that the LC rate in CRM-positive patients was 80%, compared to 93% in CRM-negative patients, and 5-year OS was 62.2% in CRM-negative patients compared to 42.2% in CRM-positive patients [16].
1.4. Extramural Venous Invasion (EMVI)
An additional imaging marker with substantial predictive value for treatment response and clinical outcomes is extramural venous invasion (EMVI). This marker demonstrates strong correlation with disease recurrence rate and, specifically, with both synchronous and metachronous systemic disease dissemination [17,18]. Positive EMVI is correlated with reduced DFS, with stage II EMVI-positive disease showing similar results to
stage III EMVI-negative disease [19]. Furthermore, a recent study suggests that EMVI is
correlated with lymph node metastases: multivariate analyses demonstrated a higher risk
of lymph node metastases in EMVI-positive patients (p = 0.0112) [20]. Such robust predictive validity has led ESGAR to recommend defining the presence of EMVI both before and
after neoadjuvant treatment [13]. This approach enables preoperative stratification of patients into low- or high-risk categories and facilitates a more personalized treatment strategy for a significant proportion of the rectal cancer patient population, with a mean EMVI
prevalence of 26%, as reported in the systematic review of Chand and colleagues, 2016
[21].
J. Clin. Med. 2026, 15, 430
4 of 20
1.5. Lateral Lymph Nodes (LLN)
Rectal cancers situated below the peritoneal reflection demonstrate a propensity to
spread laterally to lymph nodes surrounding the internal iliac and obturator vessels [22].
These lateral lymph nodes (LLN) are located outside the standard surgical plane of total
mesorectal excision (TME). In the era before neoadjuvant CRT and TME, local recurrences
(LR) occurred frequently and were often localized in the pelvis [23]. However, nowadays,
although pelvic LC shows very high rates, exceeding 90%, the absolute risk of LR is approximately 50% in the lateral compartments [8,11]. This is most likely due to LLN not yet
being appropriately treated. In 30–40% of patients with primarily enlarged LLN (>10 mm,
short axis) treated with CRT and TME, LR occurs within 5 years [8]. A recent international
cohort of 1216 patients with standardized review of all MRIs found that patients with
enlarged LLN (≥7 mm) before CRT had a 5-year LR rate of 19.5% [6]. LLN ≥ 7 mm before
CRT in the internal iliac compartment that remained >4 mm at restaging had a 5-year LLR
rate of 52.3%. Obturator LLN had a 5-year LLR risk of 17.8% when remaining > 6 mm.
Only 22% of internal iliac LLN significantly reduced in size (18 years old, possessed histologically confirmed primary rectal adenocarcinoma and no extrapelvic disease (TNM
staging as cT2-4 cN0-2). We reviewed all available medical records in both digital and
paper format for data collection.
Pre- and post-CRT staging utilized physical examination, digital rectal examination,
and chest-abdomen-pelvis CT scan. Throughout the 5 years following CRT, rectal MRI
was performed routinely.
2.2. Radiotherapy and Concomitant Chemotherapy
RT was administered using the volumetric modulated arc therapy (VMAT) technique, delivering a total dose of 45 Gy, 1.8 Gy/day, to the pelvic lymph nodes and 55 Gy,
2.2 Gy/day to the entire mesorectum in simultaneous integrated boost (SIB), corresponding to an equivalent dose at 2 Gy/fraction (EQD2) of 57.5 Gy (considering α/β = 5.06 Gy
for rectal tumor). During simulation, patients were immobilized supine and instructed to
consume 750 mL of water over 45 min to achieve appropriate bladder volume. The clinical
target volume (CTV) encompassed the primary tumor and mesorectal, presacral, and
J. Clin. Med. 2026, 15, 430
5 of 20
pelvic lymph nodes up to the L5/S1 junction. The CTV Boost was delineated to include
the primary tumor and mesorectum. The planning target volume (PTV) and PTV Boost
comprised their corresponding CTV plus an 8 mm margin in all directions. Dose specification followed the International Commission on Radiation Units and Measurements
(ICRU 50–62) report. All patients received concomitant chemotherapy: Capecitabine 825
mg/m2, twice daily for 5 days/week during radiation treatment.
2.3. Surgery
Radical surgery, including anterior resection (AR) with total mesorectal excision
(TME) or abdominoperineal resection (APR), with colorectal or colo-anal anastomosis,
was performed according to surgical assessment.
2.4. Adjuvant Chemotherapy
Adjuvant chemotherapy was administered to 47 patients (26%). Specifically, 39 patients received capecitabine and 4 capecitabine + oxaliplatin, while FOLFIRI + bevacizumab, FOLFIRI + panitumumab, FOLFOX regimens, and gemcitabine were administered to 1 patient each, respectively.
2.5. Toxicity
Radiation Therapy Oncology Group (RTOG) toxicity criteria were used to evaluate
acute RT toxicities. Routine postoperative follow-up examinations were performed every
6 months during the first 5 years following surgery, then annually. Gastrointestinal, urinary, hematological, and cutaneous symptoms were evaluated at baseline, during treatment, and at each follow-up examination. Late toxicities were reported according to the
RTOG/European Organization for Research and Treatment of Cancer (EORTC) scoring
system [26].
2.6. MRI Radiomic Features
Patients underwent MRI at two time points: before CRT initiation and between 8–10
weeks after treatment completion, defined as pre-CRT MRI and post-CRT MRI, respectively.
A multidisciplinary team consisting of an experienced radiologist, junior radiologists,
an experienced radiation oncologist, and a junior radiation oncologist performed consensus review of all available MRIs from the study population. Measurements and assessments were performed collaboratively using a consensus approach, with preliminary
evaluations by junior team members followed by review and final determinations made
by the senior radiologist and radiation oncologist when disagreements arose. All assessments were based on established international guidelines [12,13].
The following morphological characteristics were recorded: tumor extension beyond
muscularis propria (TEMP), circumferential resection margin (CRM), extramural venous
invasion (EMVI), and lateral lymph nodes (LLN).
According to national and international guidelines [12,13,27], we divided patients
into: TEMP 5 mm. Also for CRM, since it is defined as positive if there is a
distance of 1 mm or less between the pathological lesion and the mesorectal fascia [12,13],
patients were divided into: CRM 1 mm. For LLN, the indications for
pathological lymph node cut-off provided by Ogura et al. [28] and Sluckin et al. [29] were
followed: LLN > 7 mm in the pre-CRT phase and LLN > 4 mm in the post-CRT phase were
considered pathological. Finally, patients with pre-CRT LLN > 7 mm and post-CRT LLN
7 mm and postCRT LLN > 4 mm as non-responders.
J. Clin. Med. 2026, 15, 430
6 of 20
2.7. Statistical Analysis
Descriptive statistics were outlined as frequencies and percentages, while all numerical variables were reported as means and standard deviation (SD). Survival analyses for
5- and 10-year OS, DFS, and LC rates were performed using the Mantel-Cox test. OS was
defined as the time interval between surgery and death; DFS was defined as the time between surgery and the first event (recurrence and/or distant metastasis), and LC was considered as the time between surgery and locoregional recurrence. For patients in whom
none of the events occurred, the observation time interval was defined as the period from
surgery to the last follow-up visit. The Mantel–Cox test was also used to estimate 5- and
10-year OS, DFS, and LC after stratifying patients by MRI characteristic subgroups. A p <
0.05 was considered statistically significant.
2.8. AI Statement
During the preparation of this manuscript, the authors used Claude Sonnet 4.5 for
the purposes of improvement of language and of readability. The authors have reviewed
and edited the output and take full responsibility for the content of this publication.
3. Results
3.1. Patients’ Characteristics (Table 1)
A total of 134 patients were analyzed in this study. Mean patient age was 67.5 years
(range: 37–94 years); 86 patients (64%) were male with a male/female ratio of 1.8:1. Mean
Karnofsky Performance Score (KPS) was 96% (range: 70–100%).
Table 1. Patients’ characteristics.
Age, year old
Gender
KPS, %
Clinical stage T
Clinical stage N
Grade
Surgery
Pathological stage T
Pathological stage N
Median, range
Female
Median
Not available
Others
Not available
Value (n)
67 (37–94)
J. Clin. Med. 2026, 15, 430
7 of 20
Abbreviations: KPS: Karnofky performance score; AR: Anterior resection; APR: Abdominoperineal
resection; TRG: Tumor regression grade.
Rectal bleeding was the most common presenting symptom in the study population
(72%), followed by pelvic pain and/or tenesmus (26%). Most patients (82%) had cT3 tumors and, simultaneously, 51% of patients presented with cN2 lymph node stage. The
most frequent histological presentation was adenocarcinoma (90%), while the most represented histological grading was G2 (53%).
3.2. Radiotherapy and Concomitant Chemotherapy
All patients underwent radiation treatment with dose intensification up to 55 Gy,
associated with capecitabine. Treatment was interrupted in 10 patients due to toxicity.
Mean follow-up was 45 months (range 4–102).
3.3. Surgery and Adjuvant Chemotherapy
One hundred twenty-four patients (92%) underwent surgery. Anterior resection was
performed in 92 patients (74%), while 26 patients (20%) underwent abdominoperineal resection. Three patients (2.2%) with favorable clinical stage (cT2N0) and a major pathological response chose a watch-and-wait approach. Other types of surgery were performed
in five patients (3.7%), while eight patients (6%) did not undergo surgery due to poor
clinical conditions. Adjuvant chemotherapy was administered to 34 patients (26%).
3.4. Acute Toxicity
The most frequent acute toxicity reported was lower gastrointestinal: 28 patients
(21%) with grade 2 toxicity, while toxicity grade ≥ G3, such as rectal bleeding and/or severe diarrhea, was not reported in any patient. The second most encountered acute toxicity was genitourinary with a total of six patients with grade 2 toxicity. Following this,
acute skin toxicity was reported in three patients experiencing grade 2, while grade G3
toxicity (moist desquamation) was reported in two patients. No severe hematological,
neurological, or hepatic toxicities occurred (Table 2).
Table 2. Acute toxicity.
Acute Toxicity
Hematological
Abbreviations: GI: Gastro-intestinal; GU: Genito-urinary.
3.5. Late Toxicity
According to the RTOG/EORTC scale, five patients (4%) reported severe late intestinal toxicity, with bleeding requiring surgical treatment. Moderate diarrhea (more than 5
episodes per day) occurred in one patient (0.7%). Moderate pollakiuria (1 urination 5 mm were 51 pre-CRT and 20 postCRT. A CRM 7
J. Clin. Med. 2026, 15, 430
9 of 20
mm (range: 8–18 mm). Of these, 14 had regression of lymph node metastases, while in 11
persistence of lateral lymph node pathological status with lymph nodes > 4 mm (range:
5–16 mm) was observed.
4. Outcomes
Five-year OS, DFS, and LC rates were 85%, 83%, and 88%, respectively. Long-term
results at 10 years showed OS, DFS, and LC rates of 75%, 81%, and 88%, respectively.
4.1. Impact of TEMP on Outcomes
Regarding pre-CRT MRI radiological characteristics, survival outcome analyses were
conducted. Starting with TEMP, the subgroup with TEMP 5 mm (Figure 1). Ten-year outcomes were as follows: for patients with TEMP
5 mm, LC:
73%, DFS: 68%, and OS: 55%. Statistical significance was found in LC curves with p = 0.02
and DFS with p = 0.04. From analyses conducted on post-CRT MRI data, it was found that
in the group with TEMP > 5 mm, 5-year LC and DFS values of 65% and 52% were obtained,
compared to the group with TEMP 1 mm, 5-year LC, DFS, and OS rates of
91%, 80%, and 85% were recorded, as well as 10-year LC, DFS, and OS rates of 91%, 79%,
and 60%. Instead, in patients with CRM 7 mm) for (a) Disease-free survival; (b) Local control.
Regarding post-treatment analysis, in the group of patients with persistence of LLN+,
we recorded a statistically significant difference in LC with p value = 0.04. Five-year LC
results were, for post-CRT LLN+, 58% versus 87% for post-CRT LLN− (Figure 5). No statistical difference was found for post-treatment DFS.
Figure 5. Mantel–Cox test in patients with and without pathological lateral compartment lymph
nodes after neoadjuvant treatment (LLN > 4 mm) for Local control.
5. Discussion
In LARC, neoadjuvant radio-chemotherapy has the potential to determine high LC
with a good probability of obtaining better quality and life expectancy [30,31]. Yet, although neoadjuvant treatment has demonstrated its important efficacy since 2012, clinical
practice shows that patients with the same clinical stage and treated with the same
J. Clin. Med. 2026, 15, 430
12 of 20
therapeutic protocol have different outcomes. From the desire to understand which factors can influence treatment response and outcomes, this study was born with the aim of
describing our Center’s experience in treating RC and analyzing radiological factors in
our study population that in the literature have been shown to have predictive value.
The integration of MRI-based radiomic features in clinical practice represents a significant advancement in personalized medicine for RC patients. Recent literature supports
the use of multiparametric MRI radiomics for outcome prediction, with studies demonstrating that radiomic models can enhance predictive accuracy beyond traditional clinical
and imaging parameters [11,24,25,32,33].
5.1. Tumor Extension Beyond Muscularis Propria (TEMP)
The extent of mesorectal tumor invasion, in other words, TEMP, has been demonstrated as an independent risk factor in numerous studies [16,34,35]. The meta-analysis
conducted by Siddiqui et al. in 2018 observed that tumors with invasion greater than 5
mm from the muscularis propria had statistically significantly worse OS and statistically
worse DFS [14]. At the same time, in patients with less invasion beyond the muscularis
propria, higher OS was observed (p 5 mm patients DFS and OS values of 60% and 53% versus
values of 85% and 93% in patients with TEMP 5 mm and those with TEMP < 5 mm (5-year DFS: 77.6% vs. 55.2%, p <
0.001) [37]. Merkel et al. also observed a similar difference in their work, with 5-year LC,
OS, and DFS rates in T3a patients (TEMP 5 mm) [34].
In our experience, we observed a difference between the TEMP 5 mm subgroup
with values of 82%, 68%, and 66%. Furthermore, statistical significance was detected for
LC curves (p = 0.02) and DFS (p = 0.04). These findings align with recent radiomic studies
showing that quantitative assessment of tumor extension provides valuable prognostic
information beyond conventional T-staging [38].
5.2. Circumferential Resection Margin (CRM)
CRM has been defined as the most important among radiological features regarding
LC and as a predictive factor for systemic disease spread. The MERCURY trial was the
first study to analyze the reliability of MRI methodology in describing this feature and to
study its predictive value. The results were OS and DFS for all patients with CRM 1 mm patients and 72% in those with CRM 1 mm or negative CRM, LC, DFS, and OS rates
of 91%, 80%, and 85% at 5 years were recorded. Instead, in patients with CRM 4 mm on restaging MRI resulted in a 5-year LC rate
of 48% [48]. The importance of LLN as a predictive factor for LC is reiterated by two further recent studies: in 2024, the Italian study conducted by Achilli et al. recorded an LC
rate in LLN-positive patients of 91% compared to 96% in LLN-negative patients [49]; in
2022, Schaap et al. observed a higher local recurrence rate in the group with LLN compared to the group without LLN (9.8% vs. 2.5%; p = 0.056) and also recorded in patients
J. Clin. Med. 2026, 15, 430
14 of 20
without LLN a significantly better OS than that of patients with LLN detected via MRI (p
= 0.021) [50].
Recent radiomic models for predicting lateral pelvic lymph node metastasis have
demonstrated superior performance compared to conventional size-based criteria, with
AUCs reaching 0.819–0.829, and have shown potential for improving clinical decisionmaking regarding lateral lymph node dissection, as shown in the multicenter retrospective study by Yoo et al. 2025 [32] and in the retrospective study by Zhao et al. 2024 [33].
In our experience, the analysis conducted between LLN+ pre-CRT vs. LLN- pre-CRT
groups showed 5-year LC and DFS of 77% and 43% versus 87% and 78%, respectively.
The results, although not statistically significant, appear promising with p values = 0.10
and 0.20, respectively. Regarding post-treatment analysis, in the group of patients with
persistence of LLN, we recorded a statistically significant difference in LC with p value =
0.04 and with 5-year rates of 58% for post-CRT LLN+ versus 87% for post-CRT LLN-.
6. Summary of Radiomic Features and Clinical Outcomes
The data observed in our study confirmed that radiomic features may serve as indicators of clinical outcomes, showing strong agreement with the literature. The two marked
reference studies reported slightly superior performance, possibly due to their selection
of a more clinically favorable patient cohort (Table 5).
Table 5. Summary and comparisons about the analyzed MRI features among the selected studies
and our experience. * = the results of these two studies are referred only to selected LLN patients
with a better clinical presentation. Abbreviations: TEMP: Tumor Extension Beyond Muscularis Propria; CRM: Circumferential Resection Margin; EMVI: Extramural Venous Invasion; LLN: Lateral
Lymph Nodes.
Study Authors
Katsumata et al. [36]
Shin et al. [37]
Merkel et al. [34]
MERCURY-Trial [16]
Patel et al. [41]
Sun et al. [42]
Wolberick et al. [40]
Chand et al. [21]
Lee et al. [47]
Ogura et al. [48]
Achilli et al. [49]
Schaap et al. [50]
Our experience
TEMP > 5 mm
5-yr OS 5-yr DFS
CRM 5 mm, CRM 5 mm, circumferential
J. Clin. Med. 2026, 15, 430
16 of 20
resection margin (CRM) < 1 mm, presence of EMVI, and LLNs were correlated with both
lower LC and lower OS and DFS. Better understanding and knowledge of these characteristics, which are already present in the patient’s diagnostic pathway and are easy to
perform and interpret, will allow greater personalization of therapy, which can thus be
“tailored” to the patient.
Looking forward, the evolution of rectal cancer imaging will likely involve integration of conventional morphological assessment with emerging AI-driven computational
radiomics and molecular biomarkers. Such integrated approaches could combine the interpretability and clinical validation of conventional features with the high-dimensional
pattern recognition capabilities of machine learning algorithms, potentially providing
complementary prognostic information and further optimizing patient stratification.
Future prospective studies and validation in multicenter cohorts are needed to establish standardized protocols for radiomic feature assessment and integration into clinical
guidelines.
Author Contributions: Conceptualization, M.L., A.D., C.R., and D.G.; methodology, M.L., A.D.,
A.D.P. (Andrea Delli Pizzi), and D.G.; validation, A.D., C.R., A.D.P. (Andrea Delli Pizzi), and D.G.;
formal analysis, M.L.; investigation, M.L., C.R., M.D.T., M.T. (Maria Taraborrelli), M.N., M.T. (Marianna Trignani), and M.D.N.; resources, C.R., A.D.P. (Andrea Delli Pizzi), and M.D.T.; data curation,
M.L., C.R., M.D.T., and M.D.N.; writing—original draft preparation, M.L.; writing—review and editing, A.D., C.R., and D.G.; visualization, M.D.T., M.T. (Maria Taraborrelli), M.N., G.d.P., T.S., A.A.,
A.V., A.D.P. (Angelo Di Pilla), M.N., and M.T. (Marianna Trignani); supervision, A.D., C.R., M.D.N.,
and D.G. All authors have read and agreed to the published version of the manuscript.
Funding: This work was not supported by any funding. All authors have declared that no financial
support was received from any organization for the submitted work.
Institutional Review Board Statement: This study qualified as a non-pharmacological retrospective
analysis of routine clinical data. Under Italian Law (specifically Law No. 56/2024, which updated