La radiologia medica

https://doi.org/10.1007/s11547-021-01421-0

ABDOMINAL RADIOLOGY

Does restaging MRI radiomics analysis improve pathological complete

response prediction in rectal cancer patients? A prognostic model
development
Giuditta Chiloiro1 · Davide Cusumano1 · Paola de Franco2 · Jacopo Lenkowicz1 · Luca Boldrini1 ·
Davide Carano2 · Brunella Barbaro1,2 · Barbara Corvari1 · Nicola Dinapoli1 · Martina Giraffa2 ·
Elisa Meldolesi1 · Riccardo Manfredi1,2 · Vincenzo Valentini1,2 · Maria Antonietta Gambacorta1,2

Received: 11 March 2021 / Accepted: 14 October 2021

© Italian Society of Medical Radiology 2021

Abstract
Purpose Our study investigated the contribution that the application of radiomics analysis on post-treatment magnetic
resonance imaging can add to the assessments performed by an experienced disease-specific multidisciplinary tumor board
(MTB) for the prediction of pathological complete response (pCR) after neoadjuvant chemoradiotherapy (nCRT) in locally
advanced rectal cancer (LARC).
Materials and methods This analysis included consecutively retrospective LARC patients who obtained a complete or
near-complete response after nCRT and/or a pCR after surgery between January 2010 and September 2019. A three-step
radiomics features selection was performed and three models were generated: a radiomics model (rRM), a multidisciplinary
tumor board model (yMTB) and a combined model (CM). The predictive performance of models was quantified using the
receiver operating characteristic (ROC) curve, evaluating the area under curve (AUC).
Results The analysis involved 144 LARC patients; a total of 232 radiomics features were extracted from the MR images
acquired post-nCRT. The yMTB, rRM and CM predicted pCR with an AUC of 0.82, 0.73 and 0.84, respectively. ROC com-
parison was not significant (p = 0.6) between yMTB and CM.
Conclusion Radiomics analysis showed good performance in identifying complete responders, which increased when com-
bined with standard clinical evaluation; this increase was not statistically significant but did improve the prediction of clinical
response.

Keywords Rectal cancer · Neoadjuvant chemoradiation · Radiomics · Response prediction · Magnetic resonance imaging ·
Multidisciplinary tumor board

Abbreviations DRE Digital rectal examination

3D-CRT​ Conformational radiotherapy techniques DWI Diffusion-weighted imaging
5-FU 5-Fluorouracil GTV Gross tumor volume
ADC Apparent diffusion coefficient IMRT Intensity-modulated radiation therapy
AIC Akaike information criterion LARC​ Locally advanced rectal cancer
cCR Clinical complete response LC Local control
CI Confidence interval LE Local excision
CM Combined model MRF Mesorectal fascia
CR Complete response MRI Magnetic resonance imaging
CT Computed tomography MTB Multidisciplinary tumor board
DFS Disease-free survival nCRT​ Neoadjuvant chemoradiotherapy
ncCR Near complete response
NPV Negative predictive value
* Paola de Franco OS Overall survival
[email protected] pCR Pathological complete response
Extended author information available on the last page of the article rRM Radiomics model

13
Vol.:(0123456789)
 La radiologia medica

ROC Receiver operating characteristic persistence of residual deep nests of viable cancer despite
ROI Region of interest cCR is still a matter of concern [15].
RT Radiotherapy Several experiences have observed that pCR occurs in
SIB Simultaneous integrated boost 15% of cases in which all modalities indicated residual
TaMIS Transanal minimally invasive surgery tumor [4, 14, 16].
TEM Transanal endoscopic microsurgery It is known that radiomics texture features of pre-treat-
TME Total mesorectal excision ment MRI are useful in identifying potential pCR after
VMAT Volumetric modulated arc therapy nCRT [17–20] and some studies have recently reported
WW Watch-and-wait promising results even considering the variation in image
features extracted from MR images acquired before and after
nCRT [21–24] or before and during nCRT [25–27].
Introduction To the best of our knowledge, only one study compared
radiomics analysis with post-treatment MRI qualitative
A standard treatment for locally advanced rectal cancer assessment [28].
(LARC) includes neoadjuvant chemoradiotherapy (nCRT), Whether the integration of radiomics analysis can add a
followed by total mesorectal excision (TME) and adjuvant significant contribution to the clinical evaluation performed
chemotherapy if clinically indicated [1, 2]. by the radiologist or whether the information extracted from
Neoadjuvant chemoradiotherapy is associated with tumor quantitative image analysis is practically already included in
downstaging, improved local control (LC) and resectability standard of radiological evaluation, is still a matter of debate
by reducing iatrogenic tumor cell seeding during surgery, in the scientific community.
increasing free margins resection (R0) and anal sphincter The purpose of this study is to evaluate the benefits of
preservation rate in low rectal cancer [2]. adding radiomics features extracted from post-treatment
Many studies reported a pathological complete response MRI to standard morphological and functional MRI report-
(pCR) rate of 18–38% after nCRT [3–5]; this means that ing and information obtained through DRE in a retrospective
some LARC patients undergoing surgery after nCRT have cohort of patients who obtained near cCR (ncCR) or cCR
received overtreatment, followed by not negligible mortality or pCR at surgery.
and morbidity rates [6] without an overall (OS) or disease-
free survival (DFS) benefit compared with the observation-
only approach [7, 8]. Materials and methods
Therefore, in recent years, patients with clinical com-
plete response (cCR) are offered close surveillance through We retrospectively collected from our institutional registries
a watch-and-wait (WW) or rectum-sparing approach [9–11]. LARC patients who achieved cCR or ncCR on restaging by
However, the achievement of cCR does not always correlate MRI and DRE after nCRT or pCR at surgery between Janu-
with pCR: Adequate identification of pCR cases is therefore ary 2010 and September 2019. The study was approved by
mandatory, in order to identify patients who could be safely our institution’s ethics committee, and informed consent was
managed with close follow-up, avoiding TME. obtained from patients when applicable.
Evaluation of response to nCRT treatment is currently Inclusion criteria were as follows: (a) age > 18 years at
performed by digital rectal examination (DRE), post-treat- the time of diagnosis; (b) histopathologically confirmed
ment magnetic resonance imaging (MRI) and endoscopy: diagnosis of rectal adenocarcinoma; (c) locally advanced
Each of these modalities shows high level of accuracy, but disease determined by pre-treatment MRI (≥ T3 and/or
also some significant pitfalls. positive nodal status); (d) pre- and post-treatment MRI per-
A recent meta-analysis and systematic review showed that formed at our institution, using the same 1.5-T MR scanner,
the overall sensitivity of MRI was 81% (95% confidence including diffusion-weighted imaging (DWI) and high-res-
interval (CI) 67–90%), the overall specificity was 67% (95% olution T2-weighted MR images; (e) all patients who under-
CI 51–80%) and the AUC was 0.81 (95% CI 0.78–0.84) [12], went nCRT; (f) surgery performed at least 8 weeks after
whereas in another study, endoscopy predicted pCR with completion of nCRT; and (g) pathological staging available.
a sensitivity of 79%, a specificity of 86% and an AUC of All patients underwent staging by clinical examina-
0.82 [13]. The combination of DRE, MRI and endoscopy tion, endoscopy, baseline MR pelvic scans and computed
increases prediction accuracy up to 98% [14]. tomography (CT) scans of thorax, abdomen and pelvis.
However, the main limitation of current diagnostic restag- Each patient was staged according to the TNM classifica-
ing with MRI is related to its inability to distinguish between tion and discussed by the institutional disease-specific mul-
small remaining tumor foci and post-treatment fibrosis; the tidisciplinary tumor board (MTB), consisting of radiation

13
La radiologia medica

oncologists, surgeons, clinical oncologists, radiologists, antiperistaltic agent, hyoscine-N-butyl bromide (20 mg), was
pathologists and endoscopists. also administered.
The total dose of radiotherapy (RT) to the mesorectum in A sagittal localizer image was obtained for the selection
toto and selected lymphatic drainage stations, according to of transverse and coronal images. All these images were
disease stage [1, 29], was 45 Gy/1.8 Gy die. The dose to the acquired by using a T2-weighted fast spin-echo sequence
tumor and the correspondent mesorectum was 50.4 Gy with (repetition time: 2500–5000 ms; echo time: 100 ms; matrix:
a sequential boost of 5.4 Gy/die or 55 Gy delivered with a 256 × 256; echo train length: 16; number of signals acquired:
concomitant boost of 2 Gy over five weeks, or of 2.2 Gy/die 4; FOV: 24 cm; slice thickness: 3 mm). In addition, oblique
by simultaneous integrated boost (SIB). high-spatial-resolution traverse images were acquired in
Radiation treatment was performed with three-dimen- a plane orthogonal to the tumor with a T2- weighted fast
sional conformational radiotherapy (3D CRT) or intensity- spin-echo sequence (repetition time: 2500/5000 ms; inver-
modulated radiation therapy (IMRT) or volumetric modu- sion time: 100 ms; matrix: 256 × 256; echo train length: 16;
lated arc therapy (VMAT) techniques. FOV: 18 cm; slice thickness: 3 mm).
Concomitant chemotherapy was carried out with the At the post-treatment MRI examination, by integrating
administration of fluoropyrimidine (oral capecitabine DWI and high-resolution T2-weighted MR images, we con-
at a dose of 1650 mg/sqm or 5-fluorouracil 250 mg/sqm sidered a decrease in signal intensity when compared with
1–5 days) associated or not with a weekly oxaliplatin (50 mg/ the pre-treatment examination to represent response with
sqm 1, 8, 21, 28 days), according to different therapeutic fibrosis. In difficult cases at post-treatment MRI, results of
schedules [30–32], depending on the clinical presentation. the first MRI examination were used to infer signs of tumor.
Clinical restaging was assessed at least 6 weeks after the The post-treatment DICOM MR images of each patient were
end of nCRT by DRE, T2-weighted and diffusion-weighted exported from the institutional PACS system and uploaded
MRI and in selected cases by rectoscopy. Response to nCRT on a radiotherapy delineation console (Eclipse, Varian Medi-
was established by MTB. cal System™, Palo Alto, California, USA).
A cCR was defined by the absence of a palpable mass A resident radiologist and radiation oncologist deline-
at DRE and of any mucosal irregularity at endoscopy. On ated the gross tumor volume (GTV) on the axial oblique
pelvic MRI, the rectal wall appeared normal or showed only T2-weighted images (Fig. 1), obtained perpendicular to
subtle hypointense wall thickening, no suspicious involved the long axis of the rectal tumor or area after treatment,
lymph nodes, no residual hyperintense signal on b1000 including the whole tumor and excluding the lumen, stools
images or a low apparent diffusion coefficient (ADC) at the and air. In case of apparent complete response at MRI, the
previous tumor site [14, 16, 33–36]. corresponding to pre-treatment MRI hypointense area in
Near cCR was defined by the absence of a palpable mass T2-weighted imaging at the level of the primary tumor was
at DRE and deep or superficial ulcer greater than 2 cm in contoured.
diameter on endoscopy. On pelvic MRI, the rectum showed Contouring was blinded with respect to histopathological
pronounced hypointense wall thickening without isointense outcomes. All contours were subsequently revised by a radi-
signal, no positive lymph nodes and no clear areas of resid- ologist and a radiation oncologist expert in rectal cancer. The
ual hyperintense signal on b1000 images [37, 38]. DICOM files containing the MR images and the RT struc-
Surgery was performed at least 8 weeks after the end of ture files containing the GTV were imported in Moddicom, a
nCRT, and patients underwent TME or local excision (LE) dedicated platform designed for radiomics analysis [39, 40].
by transanal endoscopic microsurgery (TEM) or transanal
minimally invasive surgery (TaMIS). Image analysis
Pathological response was assessed for each patient, and
pCR was defined as the absence of viable tumor cells in the Radiomics analysis was performed considering the GTV as
primary tumor and lymph nodes at the pathological speci- region of interest (ROI) and the post-treatment MR as refer-
men (ypT0ypN0). ence image on axial oblique T2-weighted images; for each
patient, a total of 232 radiomics features were extracted,
Image acquisition and selection belonging to statistical, morphological and textural families.
As regards textural features, they were calculated consider-
Pre- and post-treatment MRI was performed using a 1.5-T ing three different kernel matrices: the co-occurrence, the
MR scanner (Horizon Advantage, GE Medical Systems, run length and the size-zone).
Milwaukee, WI, USA). All patients were imaged in the Features selection was performed considering the predic-
supine position and, to reduce potential artifacts, received tive performance at the univariate analysis using the Wil-
an enema of sonography transmission gel (60–100 cc) to dis- coxon–Mann–Whitney test. Correlation between significant
tend the rectal lumen and limit luminal air. An intramuscular variables was assessed using the Pearson’s correlation method

13
 La radiologia medica

Fig. 1  GTV outline on the axial

oblique T2-weighted images.
The red line marks the scar (a)
or scar + residual tumor (b).
GTV: gross tumor volume

[41]: Features showing a Pearson’s correlation value greater Results

than 0.75 were considered correlated.
A logistic regression model was developed to predict the The analysis was conducted on 144 LARC patients (89
treatment outcome: Optimal features were obtained using a males and 55 females) that met the inclusion criteria,
stepwise method based on the Akaike information criterion whose clinical and therapeutic characteristics are summa-
(AIC) value minimization [42]. rized in Table 1.
Finally, three predictive models for pCR outcome were Patients were clinically evaluated by a DRE at a
generated: median of 6.5 weeks (range 5.5–9) after the end of nCRT,
while post-treatment MRI was performed at a median of
i) a radiomics model (rRM) that included the radiomics 7.8 weeks (range 4.5–15).
features extracted from post-treatment MRI; In 38 patients (26%), rectoscopy was performed at
ii) an MTB model (yMTB) which classified the response restaging, while in 10 patients (7%), a second MRI was
to nCRT assessed by MTB by considering four classes: performed, at a median of 15.3 weeks (range 11.9–29.2).
absence of response to therapy (class 0), partial response All the MR images were considered appropriate from
(1), ncCR (2) and complete clinical response (3); a technical point of view, and the GTV was segmented on
iii) a combined model (CM), which is a logistic regression the post-treatment MR images according to the aforemen-
that includes selected radiomics features and yMTB val- tioned procedure: The median residual volume was 2.58 cc
ues. (range 0.1–42.35 cc).

The predictive performance of all these models was quanti-

fied using the receiver operating characteristic (ROC) curve yMTB model
and evaluating the area under the ROC curve (AUC).
The obtained ROC curves were then compared to define the All the clinical responses after nCRT were assessed at
best predictive model. the yMTB: A total of 59 patients (41%) showed cCR,
The comparison among the ROC curves was performed while 34 patients (24%) showed ncCR. At the pathologi-
using the De Long’s test [43]. A classification matrix was cal- cal restaging, pCR was observed in 117 patients (81%).
culated for each model at the Youden’s best cut-off threshold to MTB’s evaluation differs than those effectively observed
assess specificity, sensitivity and positive and negative predic- in the histopathological specimen in 54% of cases, while
tive values. Differences were considered significant when a p the concordance was 46%. Table 2 shows the concord-
value less than 0.05 was reached. All statistical analyses were ance between the yMTB response and the pathological
performed using the R software, version 3.4 (R Core Team, response.
Vienna, Austria).

13
La radiologia medica

Table 1  Clinical characteristics of the patients enrolled in the study Table 1  (continued)
Characteristics No. (%) Characteristics No. (%)

Median age (range) 65 (26–82) Residual disease 10 (7%)

Sex Unvaluable/unknown 107 (74%)
Male 89 (62%) yMTB response
Female 55 (38%) 3 (Complete response) 59 (41%)
Tumor location 2 (Major/near complete response) 34 (24%)
Middle–upper 74 (51%) 1 (Partial response) 49 (34%)
Lower 70 (49%) 0 (Non-response) 2 (1%)
Clinical T stage Median days of surgical interval 88 (55–479)
cT2 17 (12%) (range)
cT3 97 (67%) Type of surgery
cT4 30 (21%) APR 27 (19%)
Clinical N stage AR 72 (50%)
cN0 17 (12%) Other 13 (8%)
cN + 127 (88%) LE 32 (22%)
MRF positive Pathologic T stage
No 97 (67%) ypT0 117 (81%)
Yes: 46 (32%): ypT1 6 (4%)
For T 14 (10%) ypT2 16 (11%)
For N 27 (19%) ypT3 5 (3%)
For T and N 5 (3%) ypT4 0
Unvaluable/unknown 1 (< 1%) Pathological N stage
Median RT dose (range) 55 Gy (range 38.6–55 Gy) ypN0 103 (72%)
Concomitant chemotherapy ypN1 6 (4%)
With oxaliplatin: 63 (44%): ypN2 3 (2%)
Capecitabine + oxaliplatin 52 (36%) NA 32 (22%)
5-Fluorouracil + oxaliplatin 11 (8%) pCR
Without oxaliplatin: 81 (56%): Yes 117 (81%)
Capecitabine 66 (46%) No 27 (19%)
5-Fluorouracil 11 (8%) APR abdominoperineal resection; AR anterior resection; DRE digi-
Other 4 (2%) tal rectal examination; GTV gross tumor volume; HP Hartmann’s
Median MRI restaging days (range) 54 (26–119) procedure; LE local excision; MRF mesorectal fascia; MRI magnetic
resonance imaging; MTB multidisciplinary tumor board; PME partial
Clinical T restage
mesorectal excision; ROI region of interest; RT radiotherapy; TaMIS
ycT0 63 (44%) transanal minimally invasive surgery; TaTME transanal total meso-
ycT1 11 (7%) rectal excision; TEM transanal endoscopic microsurgery; TME total
ycT2 47 (33%) mesorectal excision
ycT3 19 (13%)
ycT4 4 (3%) Table 2  Concordance cCR ncCR
Clinical N restage between the yMTB response
ycN0 102 (71%) (clinical response: complete, pCR 35 (24%) 31 (22%)
near complete, partial or no
ycN + 42 (29%) No pCR 24 (17%) 3 (2%)
response) and the pathological
Average ROI yGTV-T (range; 4.95 cc (0.1–42.35 cc; 2.58 cc) response
median) cCR Clinical complete
response. ncCR Near complete
yDRE assessment response. pCR Pathological
Complete response 25 (17%) complete response
Major response 38 (26%)
Partial response 55 (38%)
yMTB predicted pCR with an AUC of 0.82 (95% CI
Unvaluable/unknown 26 (18%)
0.76–0.87), resulting in a sensitivity of 0.70 and a speci-
Endoscopy assessment at restage
ficity of 0.89.
Scar 20 (14%)
Ulcer 7 (5%)

13
 La radiologia medica

Table 3  Coefficients and the variables of the logistic regression radi- CM model
omics model
Parameter Coefficient Standard error p-value CM predicted pCR with an AUC of 0.84 (95% CI 0.77–0.91)
with a sensitivity of 0.68 and a specificity of 0.93. Table 4
Intercept 31.84 10.45 0.002
shows sensitivity and specificity of the models.
F_cm.2.5Dmerged.joint.max − 367.99 114.28 0.001
Figure 2 reports the ROC curves of the three models elab-
F_rlm.rlnu.norm − 30.92 10.87 0.004
orated. The Delong’s test reported no statistical significance
(p = 0.6) between the two predictive models showing the
highest predictive performances, which are the yMTB and
Table 4  Predictive performances of the models object of this study the CM model. Statistical significance was instead observed
Model AUC (95% CI) Best threshold Sensitivity Specificity between yMTB and rRM (p = 0.08) and between CM model
and rRM (p = 0.004).
rRM 0.73 (0.63–0.84) 0.834 0.67 0.77
yMTB 0.82 (0.76–0.87) 0.760 0.70 0.89
CM 0.84 (0.77–0.91) 0.912 0.68 0.93 Discussion
CM Combined model. rRM Radiomics model. yMTB Multidiscipli-
nary tumor board model In this study, we investigated the ability of post-treatment
T2-weighted MRI radiomics features to predict pathologi-
cal status in a retrospective cohort of LARC patients who
rRM model achieved near cCR or cCR or pCR after surgery.
The importance of the multidisciplinary team in the
Two radiomics features showed statistical significance (p staging and restaging in rectal cancer is widely known
< 0.05) in discriminating cCR at the univariate analysis: [44]; a combination of textural and morphological radiom-
the textural joint maximum considering the co-occurrence ics descriptors was combined with clinical information in
matrix and the non-uniformity considering the run length order to assess the advantage of implementing radiomics
matrix. Table 3 contains the coefficients and the variables on the assessment of MTB. This study represents, to our
of the logistic regression radiomics model (rRM).The rRM knowledge, the first experience that quantified the predictive
predicted pCR with a performance of AUC of 0.73 (95% power of a MTB and of a CM at the restaging to predict pCR
confidence interval (CI) 0.63–0.84), a sensitivity of 0.67 and after nCRT in LARC patients.
a specificity of 0.77 at the best threshold which corresponds In this subset of patients, our analysis did not show a
to 0.834. statistical benefit of radiomics features compared to a multi-
examination clinical evaluation based on T2-weighted and
DWI MRI, DRE and endoscopy.
The pure radiomics analysis reported a moderate perfor-
mance in identifying complete responders (rRM-AUC of
0.73), while the current clinical and radiological evaluation

Fig. 2  ROC curves representing the performance of the three models to predict the pCR outcome. a Radiomics model (rRM); b multidiscipli-
nary tumor board model (yMTB); and c combined model (CM)

13
La radiologia medica

performed by an experienced MTB showed superior per- and identified radiomics textural and morphological
formance with an AUC of 0.82, which is better than some descriptors of the rectal wall as main indicators to dis-
literature data reported using MRI only (AUC of 0.81, [12]) criminate between the low (ypT0-2) and high (ypT3-4)
and the combination of endoscopy and MRI (AUC 0.77, pathological stages of cancer [48].
[13]). The limited number of studies in the literature based
As regards the radiomics model, it is composed by two on post-treatment MRI is partially due to the difficulty in
textural features based on run length and correlation matrix: contouring the residual volume after nCRT, which is often
Although a direct interpretation of such parameters is very very small, as also reported in our experience, where the
challenging, the significance of such features indicates the median residual volume was 2.58 cc (range 0.1–42.35 cc).
presence of a significant difference in the spatial distribution One of the limiting factors of radiomics in clinical prac-
of the gray levels among the different categories of treat- tice is represented by the reduced interpretability of the
ment response, as already observed in similar experiences used features, which leads to a low clinical reliability of
obtained analyzing MR images [44–46]. the elaborated models [49–52].
The addition of radiomics to the assessment of an expe- However, it should be kept in mind that this study suf-
rienced MTB increased the performance of the model, but fers from different limiting factors.
this increase was not statistically significant (yMTB—AUC First of all, the study had a retrospective design: Patients
0.82 versus CM—AUC 0.84, p = 0.6, at the DeLong’s test). were included in different time periods (between January
The non-statistical significance could be due to the small 2010 and September 2019) and this may have inserted
size of the contoured volumes, formed by the post-treatment differences related to staging, treatment techniques and
scar only or by the scar and the residual disease when pre- therapeutic schedules, as changes occurred over time.
sent, or it could be due to the absence of techniques that In addition, the imaging was performed using a 1.5-T MR
allow to identify the most predictive features or their correct scanner and using a scanner with an higher field strength
interpretation. might result in the extraction of more predictable features,
Through a quantification of the MTB performance, we as also reported in some experiences dealing this topic [53].
observed that MTB’s evaluation differs than those effectively The radiomics analysis was not performed on DWI and
observed in the histopathological specimen in 54% of cases, pre-treatment MRI, so it could be of interest to compare
while the concordance was 46%. the performance of the MTB with known models devel-
Some studies, in fact, showed that discrepancies between oped in rectal cancer. Moreover, the study is mono-insti-
MRI and pathological findings were mainly caused by mis- tutional and does not benefit from an external validation
interpretation of fibrosis, as well as inflammatory cell infil- cohort (TRIPOD 1, [54]).
tration, pure mucin, oedematous mucosa and submucosa Despite the acknowledged limits, this study has demon-
adjacent to the tumor and muscularis propria could be mis- strated that radiomics performed on post-treatment images
interpreted as residual tumor [47]. does not overcome the evaluation of the disease-specific
Therefore, there is still a gap in pCR prediction and these MTB. Radiomics analysis cannot be an alternative to tradi-
data stimulate further studies. tional predictive tools but can be used as an additional tool
In the literature, many studies evaluated the predictive that improves the performance of existing ones, represent-
power of radiomics analysis both at the time of clinical stag- ing “further step forward” of pCR prediction.
ing and during treatment and as variations between pre- and The integration of more comprehensive radiomics anal-
post-treatment [21–24] and, using pre- and post-treatment ysis, such as the ones performed on pre-treatment MRI
MRI data, several radiomics models have been proposed with a Delta Radiomics approach, may further improve the
with the aim of identifying a subgroup of LARC patients discriminative power of the MTB opening new promising
where pCR could be obtained after nCRT. perspectives on the use of artificial intelligence as decision
Few studies examined radiomics analysis based exclu- support tool in clinical practice.
sively on post-treatment images in LARC. A recent study
compared the qualitative information provided by a post-
treatment MRI with the quantitative information provided Funding This research received no external funding.
by radiomics analysis performed on the same image modal-
Data availability Data are available on request due to restrictions, eg,
ity. Radiomics showed higher diagnostic performance in privacy or ethical.
the analysis of T2-weighted imaging and DWI separately,
with a higher specificity and PPV when compared with Declarations
T2-weighted MRI and DWI [28].
Another recent study analyzed the T2-weighted post- Conflict of interest The authors declare no conflict of interest.
treatment MR images to predict the pathological response

13
 La radiologia medica

Ethical approval The study was conducted according to the guidelines classification predicts early tumor regrowth after nonoperative
of the Declaration of Helsinki and approved by the Ethics Commit- management in distal rectal cancer after extended neoadjuvant
tee of Fondazione Policlinico Universitario Agostino Gemelli IRCCS chemoradiation and initial complete clinical response. Dis Colon
(protocol code 3699, approved on December 22, 2020). Rectum 60:586–594. https://​doi.​org/​10.​1097/​DCR.​00000​00000​
000830
Informed consent Where applicable, informed consent was obtained. 10. Bushati M, Pucciarelli S, Gennaro N, Maretto I, Toppan P, Perin
A, Urso EDL, Bagatella A, Spolverato G (2019) Local excision
in rectal cancer patients with major or complete clinical response
after neoadjuvant therapy: a case-matched study. Int J Colorectal
Dis 34:2129–2136. https://​doi.​org/​10.​1007/​s00384-​019-​03420-0
References 11. Chiloiro G, Meldolesi E, Giraffa M, Dino N, Barbaro B, Coco C,
Corvari B, De Franco P, Ugo DD, Alfieri S, Manfredi R, Valen-
1. Benson AB, Al-Hawary MM, Arain MA, Chen Y-J, Ciombor KK, tini V, Antonietta M (2021) Clinical and translational radiation
Cohen S, Deming D, Garrido-Laguna I, Grem JL, Buffett P, Hoffe oncology could the conservative approach be considered safe in
S, Hubbard J, Hunt S, Kirilcuk N, Krishnamurthi S, Messersmith the treatment of locally advanced rectal cancer in case of a clinical
WA, Meyerhardt J, Brigham D-F, Miller CD, Nurkin S, Overman near-complete or complete response ? A retrospective analysis.
MJ, Parikh A, Patel H, Pedersen K, Saltz L, Schneider C, Shibata Clin Transl Radiat Oncol 28:1–9. https://​doi.​org/​10.​1016/j.​ctro.​
D, Skibber JM, Sofocleous CT, Stoffel EM, Stotsky-Himelfarb E, 2021.​02.​009
Willett CG, Gurski L, Johnson-Chilla A (2020) NCCN Guidelines 12. Meta-analysis CA, Review S, Wei M, Zhao Z, Wang J (2020)
Insights: Rectal Cancer, Version 6.2020. J Natl Compr Canc Netw The diagnostic accuracy of magnetic resonance imaging in restag-
18(7):806–815. https://​doi.​org/​10.​6004/​jnccn.​2020.​0032 ing of rectal cancer after preoperative. J Comput Assist Tomogr
2. Li Y, Wang J, Ma X, Tan L, Yan Y, Xue C, Hui B, Liu R, Ma H 44:102–110. https://​doi.​org/​10.​1097/​RCT.​00000​00000​000964
(2016) A review of neoadjuvant chemoradiotherapy for locally 13. Cho MS, Kim H, Han YD, Hur H, Min BS, Baik SH, Cheon
advanced rectal cancer. Int J Biol Sci. https://d​ oi.o​ rg/1​ 0.7​ 150/i​ jbs.​ JH, Lim JS, Lee KY, Kim NK (2019) Endoscopy and magnetic
15438 resonance imaging- based prediction of ypT stage in patients with
3. Monson JRT, Weiser MR, Buie WD, Chang GJ, Rafferty JF rectal cancer who received chemoradiotherapy: results from a pro-
(2013) Practice parameters for the management of rectal cancer spective study of 110 patients. Medicine 98(35):e16614
(revised). Dis Colon Rectum 56:535–550. https://d​ oi.o​ rg/1​ 0.1​ 097/​ 14. Maas M, Lambregts DMJ, Nelemans PJ, Heijnen LA, Martens
DCR.​0b013​e3182​8cb66c MH, Leijtens JWA, Sosef M, Hulsewé KWE, Hoff C, Breukink
4. Yuval JB, Thompson HM, Garcia-Aguilar J (2020) Organ pres- SO, Stassen L, Beets-Tan RGH, Beets GL (2015) Assessment of
ervation in rectal cancer. J Gastrointest Surg. https://​doi.​org/​10.​ clinical complete response after chemoradiation for rectal cancer
1007/​s11605-​020-​04583-w with digital rectal examination, endoscopy, and MRI: selection for
5. Sanghera P, Wong DWY, McConkey CC, Geh JI, Hartley A organ-saving treatment. Ann Surg Oncol 22:3873–3880. https://​
(2008) Chemoradiotherapy for rectal cancer: an updated analysis doi.​org/​10.​1245/​s10434-​015-​4687-9
of factors affecting pathological response. Clin Oncol 20:176– 15. Prezzi D, Goh V (2016) Rectal cancer magnetic resonance imag-
183. https://​doi.​org/​10.​1016/j.​clon.​2007.​11.​013 ing: imaging beyond morphology. Clin Oncol 28:83–92. https://​
6. Smith FM, Waldron D, Winter DC (2010) Rectum-conserving doi.​org/​10.​1016/j.​clon.​2015.​10.​010
surgery in the era of chemoradiotherapy. Br J Surg 97:1752–1764. 16. Habr-Gama A, Perez RO, Wynn G, Marks J, Kessler H, Gama-
https://​doi.​org/​10.​1002/​bjs.​7251 Rodrigues J (2010) Complete clinical response after neoadjuvant
7. Habr-Gama A, Perez RO, Nadalin W, Sabbaga J, Ribeiro U, E chemoradiation therapy for distal rectal cancer: characterization
Sousa AHS, Campos FG, Kiss DR, Gama-Rodrigues J, Michelassi of clinical and endoscopic findings for standardization. Dis Colon
F, Rothenberger DA, Habr-Gama A (2004) Operative versus Rectum 53:1692–1698. https://d​ oi.o​ rg/1​ 0.1​ 007/D
​ CR.0​ b013e​ 3181​
nonoperative treatment for stage 0 distal rectal cancer following f42b89
chemoradiation therapy: long-term results. Ann Surg 240:711– 17. Antunes JT, Ofshteyn A, Bera K, Wang EY, Brady JT, Willis JE,
718. https://​doi.​org/​10.​1097/​01.​sla.​00001​41194.​27992.​32 Friedman KA, Marderstein EL, Kalady MF, Stein SL, Purysko
8. van der Valk MJM, Hilling DE, Bastiaannet E, Kranenbarg EMK, AS, Paspulati R, Gollamudi J, Madabhushi A, Viswanath SE
Beets GL, Figueiredo NL, Habr-Gama A, Perez RO, Renehan AG, (2020) Radiomic features of primary rectal cancers on base-
van de Velde CJH, Ahlberg M, Appelt A, Asoglu O, Bär MT, line T2-weighted MRI are associated with pathologic complete
Barroca R, Beets-Tan RGH, Belgers EHJ, Bosker RJI, Breukink response to neoadjuvant chemoradiation: a multisite study. J Magn
SO, Bujko K, Carvalho C, Cunningham C, Creavin B, D’Hoore Reson Imaging. https://​doi.​org/​10.​1002/​jmri.​27140
A, Gérard JP, Gollins S, Hoff C, Holman FA, Hupkens BJP, Iseas 18. Dinapoli N, Barbaro B, Gatta R, Chiloiro G, Casà C, Masciocchi
S, Jakobsen A, Keshvari A, Koopal SA, Kusters M, Langheinrich C, Damiani A, Boldrini L, Gambacorta MA, Dezio M, Mattiucci
M, Leijtens JWA, Maas M, Malcomson L, Mamedli ZZ, Martling GC, Balducci M, van Soest J, Dekker A, Lambin P, Fiorino C,
A, Matzel KE, Melenhorst J, Morici ML, Murad-Regadas SM, Sini C, De Cobelli F, Di Muzio N, Gumina C, Passoni P, Manfredi
O’Dwyer ST, Peeters KCMJ, Rosa I, Rossi G, Rutten HJT, Loria R, Valentini V (2018) Magnetic resonance, vendor-independent,
FS, van der Sande ME, Julião GPS, Saunders M, Myint AS, van intensity histogram analysis predicting pathologic complete
der Sluis H, Schiappa R, Scott N, Stoot JHMB, Talsma AK, Ter- response after radiochemotherapy of rectal cancer. Int J Radiat
rasson I, Tokmak H, Vaccaro CA, Vahrmeijer AL, Wasowicz DK, Oncol Biol Phys 102:765–774. https://​doi.​org/​10.​1016/j.​ijrobp.​
Westreenen HL, Winter DC, Wolthuis AM, Zimmerman DDE 2018.​04.​065
(2018) Long-term outcomes of clinical complete responders after 19. Shaish H, Aukerman A, Vanguri R, Spinelli A, Armenta P, Jam-
neoadjuvant treatment for rectal cancer in the international watch bawalikar S, Makkar J, Bentley-Hibbert S, Del Portillo A, Kiran
& wait database (IWWD): an international multicentre registry R, Monti L, Bonifacio C, Kirienko M, Gardner KL, Schwartz
study. Lancet 391:2537–2545. https://​doi.​org/​10.​1016/​S0140-​ L, Keller D (2020) Radiomics of MRI for pretreatment predic-
6736(18)​31078-X tion of pathologic complete response, tumor regression grade,
9. Habr-Gama A, Julião GPS, Gama-Rodrigues J, Vailati BB, Ortega and neoadjuvant rectal score in patients with locally advanced
C, Fernandez LM, Araújo SEA, Perez RO (2017) Baseline T rectal cancer undergoing neoadjuvant chemoradiation: an

13
La radiologia medica

international multicenter study. Eur Radiol. https://​doi.​org/​10.​ 31. Rödel C, Graeven U, Fietkau R, Hohenberger W, Hothorn T,
1007/​s00330-​020-​06968-6 Arnold D, Hofheinz RD, Ghadimi M, Wolff HA, Lang-Welzen-
20. Cusumano D, Dinapoli N, Boldrini L, Chiloiro G, Gatta R, bach M, Raab HR, Wittekind C, Ströbel P, Staib L, Wilhelm M,
Masciocchi C, Lenkowicz J, Casà C, Damiani A, Azario L, Grabenbauer GG, Hoffmanns H, Lindemann F, Schlenska-Lange
Van Soest J, Dekker A, Lambin P, De Spirito M, Valentini V A, Folprecht G, Sauer R, Liersch T (2015) Oxaliplatin added to
(2018) Fractal-based radiomic approach to predict complete fluorouracil-based preoperative chemoradiotherapy and postop-
pathological response after chemo-radiotherapy in rectal can- erative chemotherapy of locally advanced rectal cancer (the Ger-
cer. Radiol Medica 123:286–295. https://​d oi.​o rg/​1 0.​1 007/​ man CAO/ARO/AIO-04 study): Final results of the multicentre,
s11547-​017-​0838-3 open-label, randomised, phase 3 trial. Lancet Oncol 16:979–989.
21. Liu Z, Zhang XY, Shi YJ, Wang L, Zhu HT, Tang Z, Wang S, https://​doi.​org/​10.​1016/​S1470-​2045(15)​00159-X
Li XT, Tian J, Sun YS (2017) Radiomics analysis for evalua- 32. Comella P (2007) A review of the role of capecitabine in the treat-
tion of pathological complete response to neoadjuvant chemo- ment of colorectal cancer. Ther Clin Risk Manag 3:421–431
radiotherapy in locally advanced rectal cancer. Clin Cancer Res 33. Bhoday J, Smith F, Siddiqui MR, Balyasnikova S, Swift RI, Perez
23:7253–7262. https://​doi.​org/​10.​1158/​1078-​0432.​CCR-​17-​1038 R, Habr-Gama A, Brown G (2016) Magnetic resonance tumor
22. Li Y, Liu W, Pei Q, Zhao L, Güngör C, Zhu H, Song X, Li C, regression grade and residual mucosal abnormality as predictors
Zhou Z, Xu Y, Wang D, Tan F, Yang P, Pei H (2019) Predicting for pathological complete response in rectal cancer postneoadju-
pathological complete response by comparing MRI-based radiom- vant chemoradiotherapy. Dis Colon Rectum 59:925–933. https://​
ics pre- and postneoadjuvant radiotherapy for locally advanced doi.​org/​10.​1097/​DCR.​00000​00000​000667
rectal cancer. Cancer Med 8:7244–7252. https://​doi.​org/​10.​1002/​ 34. Akce M, El-Rayes BF (2019) Nonsurgical management of rectal
cam4.​2636 cancer. J Oncol Pract 15:123–131. https://​doi.​org/​10.​1200/​jop.​18.​
23. Bulens P, Couwenberg A, Intven M, Debucquoy A, Vandecav- 00769
eye V, Van Cutsem E, D’Hoore A, Wolthuis A, Mukherjee P, 35. Lambregts DMJ, Vandecaveye V, Barbaro B, Bakers FCH, Lam-
Gevaert O, Haustermans K (2020) Predicting the tumor response brecht M, Maas M, Haustermans K, Valentini V, Beets GL, Beets-
to chemoradiotherapy for rectal cancer: model development and Tan RGH (2011) Diffusion-weighted MRI for selection of com-
external validation using MRI radiomics. Radiother Oncol. https://​ plete responders after chemoradiation for locally advanced rectal
doi.​org/​10.​1016/j.​radonc.​2019.​07.​033 cancer: a multicenter study. Ann Surg Onclol. https://​doi.​org/​10.​
24. Jeon SH, Song C, Chie EK, Kim B, Kim YH, Chang W, Lee YJ, 1245/​s10434-​011-​1607-5
Chung JH, Chung JB, Lee KW, Kang SB, Kim JS (2019) Delta- 36. Beets-Tan RGH, Lambregts DMJ, Maas M, Bipat S, Barbaro B,
radiomics signature predicts treatment outcomes after preopera- Curvo-Semedo L, Fenlon HM, Gollub MJ, Gourtsoyianni S, Hal-
tive chemoradiotherapy and surgery in rectal cancer. Radiat Oncol ligan S, Hoeffel C, Kim SH, Laghi A, Maier A, Rafaelsen SR,
14:1–10. https://​doi.​org/​10.​1186/​s13014-​019-​1246-8 Stoker J, Taylor SA, Torkzad MR, Blomqvist L (2018) Magnetic
25. Boldrini L, Cusumano D, Chiloiro G, Casà C, Masciocchi C, resonance imaging for clinical management of rectal cancer:
Lenkowicz J, Cellini F, Dinapoli N, Azario L, Teodoli S, Gam- updated recommendations from the 2016 European society of
bacorta MA, De Spirito M, Valentini V (2019) Delta radiomics gastrointestinal and abdominal radiology (ESGAR) consensus
for rectal cancer response prediction with hybrid 0.35 T magnetic meeting. Eur Radiol 28:1465–1475. https://​doi.​org/​10.​1007/​
resonance-guided radiotherapy (MRgRT): a hypothesis-generating s00330-​017-​5026-2
study for an innovative personalized medicine approach. Radiol 37. Hupkens BJP, Maas M, Martens MH, van der Sande ME, Lam-
Medica 124:145–153. https://d​ oi.o​ rg/1​ 0.1​ 007/s​ 11547-0​ 18-0​ 951-y bregts DMJ, Breukink SO, Melenhorst J, Houwers JB, Hoff C,
26. Fiorino C, Gumina C, Passoni P, Palmisano A, Broggi S, Cat- Sosef MN, Leijtens JWA, Berbee M, Beets-Tan RGH, Beets GL
taneo GM, Di Chiara A, Esposito A, Mori M, Raso R, Ronzoni (2018) Organ preservation in rectal cancer after chemoradiation:
M, Rosati R, Slim N, De Cobelli F, Calandrino R, Di Muzio NG should we extend the observation period in patients with a clinical
(2018) A TCP-based early regression index predicts the pathologi- near-complete response? Ann Surg Oncol 25:197–203. https://d​ oi.​
cal response in neo-adjuvant radio-chemotherapy of rectal cancer. org/​10.​1245/​s10434-​017-​6213-8
Radiother Oncol 128:564–568. https://​doi.​org/​10.​1016/j.​radonc.​ 38. Cianci R, Cristel G, Agostini A, Ambrosini R, Calistri L, Petralia
2018.​06.​019 G, Colagrande S (2020) MRI for rectal cancer primary staging and
27. Cusumano D, Boldrini L, Yadav P, Yu G, Musurunu B, Chiloiro restaging after neoadjuvant chemoradiation therapy: how to do it
G, Piras A, Lenkowicz J, Placidi L, Broggi S, Romano A, Mori during daily clinical practice. Eur J Radiol 131:109238. https://​
M, Barbaro B, Azario L, Gambacorta MA, De Spirito M, Bassetti doi.​org/​10.​1016/j.​ejrad.​2020.​109238
MF, Yang Y, Fiorino C, Valentini V (2020) External validation of 39. Gatta R, Vallati M, Dinapoli N, Masciocchi C, Lenkowicz J,
early regression index (ERITCP) as predictor of pathologic com- Cusumano D, Casá C, Farchione A, Damiani A, van Soest J, Dek-
plete response in rectal cancer using magnetic resonance-guided ker A, Valentini V (2019) Towards a modular decision support
radiation therapy. Int J Radiat Oncol Biol Phys. https://d​ oi.o​ rg/1​ 0.​ system for radiomics: a case study on rectal cancer. Artif Intell
1016/j.​ijrobp.​2020.​07.​2323 Med 96:145–153. https://​doi.​org/​10.​1016/j.​artmed.​2018.​09.​003
28. Horvat N, Veeraraghavan H, Khan M, Blazic I, Zheng J, Capanu 40. Dinapoli N, Alitto AR, Vallati M, Gatta R, Autorino R, Boldrini L,
M, Sala E, Garcia-Aguilar J, Gollub MJ, Petkovska I (2018) Mr Damiani A, Valentini V (2015) Moddicom: A complete and eas-
imaging of rectal cancer: radiomics analysis to assess treatment ily accessible library for prognostic evaluations relying on image
response after neoadjuvant therapy. Radiology 287:833–843. features. Proc Annu Int Conf IEEE Eng Med Biol Soc EMBS
https://​doi.​org/​10.​1148/​radiol.​20181​72300 2015-Novem. Doi: https://​doi.​org/​10.​1109/​EMBC.​2015.​73184​76
29. Valentini V, Antonietta M, Barbaro B, Chiloiro G, Coco C, Das 41. Thompson JRT (1998) An introduction to error analysis: the study
P, Fanfani F, Joye I, Kachnic L, Maingon P, Marijnen C, Ngan S, of uncertainties in physical measurements. Phys Today 51(1):57.
Haustermans K (2016) International consensus guidelines on clin- https://​doi.​org/​10.​1063/1.​882103
ical target volume delineation in rectal cancer. Radiother Oncol 42. Aike HAI (1974) A new look at the statistical model identifi-
120:195–201. https://​doi.​org/​10.​1016/j.​radonc.​2016.​07.​017 cation. Trans Autom Control 19(6):716–723. https://​doi.​org/​10.​
30. Hong TS, Kachnic LA (2007) Preoperative chemoradiotherapy 1109/​TAC.​1974.​11007​05
in the management of localized rectal cancer: the new standard. 43. Delong ER, Delong DM, Clarke-pearson DL, Delong ER,
Gastrointest Cancer Res 1:49–56 Carolina N (2013) Comparing the areas under two or more

13
 La radiologia medica

correlated receiver operating characteristic curves: a nonpara- 49. Lohmann P, Bousabarah K, Hoevels M, Treuer H (2020) Radi-
metric approach. Biometrics 44:837–845 omics in radiation oncology—basics, methods, and limita-
44. Zwanenburg A, Vallières M, Abdalah MA, Aerts HJWL, tions. Strahlentherapie Und Onkol. https://​d oi.​o rg/​1 0.​1 007/​
Andrearczyk V, Apte A, Ashrafinia S, Bakas S, Beukinga RJ, Boe- s00066-​020-​01663-3
llaard R, Bogowicz M, Boldrini L, Buvat I, Cook GJR, Davatzikos 50. Rogers W, Thulasi Seetha S, Refaee TAG, Lieverse RIY, Gran-
C, Depeursinge A, Desseroit MC, Dinapoli N, Dinh CV, Eche- zier RWY, Ibrahim A, Keek SA, Sanduleanu S, Primakov SP,
garay S, Naqa IE (2020) The image biomarker standardisation Beuque MPL, Marcus D, van der Wiel AMA, Zerka F, Oberije
initiative. Radiology 295:328–338. https://d​ oi.o​ rg/1​ 0.1​ 148/r​ adiol.​ CJG, van Timmeren JE, Woodruff HC, Lambin P (2020) Radi-
20201​91145 omics: from qualitative to quantitative imaging. Br J Radiol.
45. Nardone V, Reginelli A, Scala F, Carbone SF, Mazzei MA, 93(1108):20190948. https://​doi.​org/​10.​1259/​bjr.​20190​948
Sebaste L, Carfagno T, Battaglia G, Pastina P, Correale P, Tini P, 51. Bibault JE, Xing L, Giraud P, El Ayachy R, Giraud N, Decazes P,
Pellino G, Cappabianca S, Pirtoli L (2019) Magnetic-resonance- Burgun A (2020) Radiomics: a primer for the radiation oncolo-
imaging texture analysis predicts early progression in rectal cancer gist. Cancer/Radiotherapie 24:403–410. https://d​ oi.​org/​10.​1016/j.​
patients undergoing neoadjuvant chemoradiation. Gastroenterol canrad.​2020.​01.​011
Res Pract 2019:1–9 52. Kumar V, Gu Y, Basu S, Berglund A, Eschrich SA, Schabath MB,
46. Cusumano VVD, Boldrini L, Yadav P, Yu G, Musurunu B, Forster K, Aerts HJ, Dekker A, Fenstermacher D, Goldgof DB,
Chiloiro G, Piras A, Lenkowicz J, Placidi L, Romano A, De Luca Hall LO, Lambin P, Balagurunathan Y, Gatenby RA, Gillies RJ
V, Votta C, Barbaro B, Gambacorta MA, Bassetti MF, Yang Y (2012) Radiomics: the process and the challenges. Magn Reson
(2021) Delta radiomics for rectal cancer response prediction using Imaging. 30(9):1234–1248. https://d​ oi.o​ rg/1​ 0.1​ 016/j.m
​ ri.2​ 012.0​ 6.​
low field magnetic resonance guided radiotherapy: an external 010
validation. Physuca Medica 84:186–191. https://d​ oi.o​ rg/1​ 0.1​ 016/j.​ 53. Cusumano D, Meijer G, Lenkowicz J, Chiloiro G, Boldrini L
ejmp.​2021.​03.​038 (2020) A field strength independent MR radiomics model to pre-
47. Jia X, Zhang Y, Wang Y, Feng C, Shen D, Ye Y, Hong N (2019) dict pathological complete response in locally advanced rectal
MRI for Restaging locally advanced rectal cancer: detailed analy- cancer Laplacian of Gaussian. Radiol Med. https://​doi.​org/​10.​
sis of discrepancies with the pathologic reference standard. AJR 1007/​s11547-​020-​01266-z
Am J Roentgenol. 213(5):1081–1090. https://​doi.​org/​10.​2214/​ 54. Collins GS, Reitsma JB, Altman DG (2015) Transparent report-
AJR.​19.​21383 ing of a multivariable prediction model for individual prognosis
48. Study M, Alvarez-Jimenez C, Antunes JT, Talasila N, Bera K, or diagnosis (TRIPOD): the TRIPOD statement. BMJ 7594:1–9.
Brady JT, Gollamudi J, Marderstein E, Kalady MF, Purysko A, https://​doi.​org/​10.​1136/​bmj.​g7594
Willis JE, Stein S, Friedman K, Paspulati R, Delaney CP, Romero
E, Madabhushi A (2020) Radiomic texture and shape descriptors Publisher's Note Springer Nature remains neutral with regard to
of the rectal environment on post-chemoradiation T2-weighted jurisdictional claims in published maps and institutional affiliations.
MRI are associated with pathologic tumor stage regression in rec-
tal cancers: a retrospective. Cancers (Basel) 12:2027. https://​doi.​
org/​10.​3390/​cance​rs120​82027

Authors and Affiliations

Giuditta Chiloiro1 · Davide Cusumano1 · Paola de Franco2 · Jacopo Lenkowicz1 · Luca Boldrini1 ·
Davide Carano2 · Brunella Barbaro1,2 · Barbara Corvari1 · Nicola Dinapoli1 · Martina Giraffa2 ·
Elisa Meldolesi1 · Riccardo Manfredi1,2 · Vincenzo Valentini1,2 · Maria Antonietta Gambacorta1,2

Giuditta Chiloiro Martina Giraffa

[email protected] [email protected]
Davide Cusumano Elisa Meldolesi
[email protected] [email protected]
Jacopo Lenkowicz Riccardo Manfredi
[email protected] [email protected]
Luca Boldrini Vincenzo Valentini
[email protected] [email protected]
Davide Carano Maria Antonietta Gambacorta
[email protected] [email protected]
Brunella Barbaro 1
Fondazione Policlinico Universitario A. Gemelli IRCCS,
[email protected]
00168 Roma, Italy
Barbara Corvari 2
Università Cattolica del Sacro Cuore, Largo Francesco Vito,
[email protected]
1, 00168 Roma, Italy
Nicola Dinapoli
[email protected]

13