Imaging biomarkers for evaluating tumor response: RECIST and beyond

Response Evaluation Criteria in Solid Tumors (RECIST) is the gold standard for assessment of treatment response in solid tumors. Morphologic change of tumor size evaluated by RECIST is often correlated with survival length and has been considered as a surrogate endpoint of therapeutic efficacy. However, the detection of morphologic change alone may not be sufficient for assessing response to new anti-cancer medication in all solid tumors. During the past fifteen years, several molecular-targeted therapies and immunotherapies have emerged in cancer treatment which work by disrupting signaling pathways and inhibited cell growth. Tumor necrosis or lack of tumor progression is associated with a good therapeutic response even in the absence of tumor shrinkage. Therefore, the use of unmodified RECIST criteria to estimate morphological changes of tumor alone may not be sufficient to estimate tumor response for these new anti-cancer drugs. Several studies have reported the low reliability of RECIST in evaluating treatment response in different tumors such as hepatocellular carcinoma, lung cancer, prostate cancer, brain glioma, bone metastasis, and lymphoma. There is an increased need for new medical imaging biomarkers, considering the changes in tumor viability, metabolic activity, and attenuation, which are related to early tumor response. Promising imaging techniques, beyond RECIST, include dynamic contrast-enhanced computed tomography (CT) or magnetic resonance imaging (MRI), diffusion-weight imaging (DWI), magnetic resonance spectroscopy (MRS), and 18 F-fluorodeoxyglucose (FDG) positron emission tomography (PET). This review outlines the current RECIST with their limitations and the new emerging concepts of imaging biomarkers in oncology.


Introduction
Tumor response assessment is important in clinical trials and standard cancer treatments. Both tumor regression and disease progression detected by imaging biomarkers are important endpoints. Imaging biomarkers can detect the subtle changes in physiology and pathology before their clinical detection and thus act as surrogate endpoints reducing time and resources used in cancer clinical trials [1]. Besides, imaging biomarkers are also used as predictive classifiers to assist in selecting appropriate candidates for particular treatments [2]. Conventionally, morphologic change in tumor size is usually related to survival length and has been considered as a surrogate endpoint of therapeutic efficacy by the World Health Organization (WHO) criteria and Response Evaluation Criteria in Solid Tumors (RECIST 1.0 and 1.1 versions) ( Table 1). Currently, RECIST 1.1 is the gold standard for the assessment of treatment response in solid tumors [3,4]. Traditional chemotherapies are cytotoxic and act primarily by eliminating neoplastic cells. Therefore, a change in tumor size indicates a change in the number of neoplastic cells and has thus evolved into a radiologic biomarker of treatment response [5]. However, unlike cytotoxic drugs, molecular-targeted therapy, and immunotherapy, which have emerged in the past 15 years, can effectively interfere with signaling pathways and thereby inhibit cell growth. For these new treatments, tumor necrosis or lack of tumor progression may be associated with an improvement in outcome, even in the absence of major shrinkage of tumors [6]. Thus, the assessment of morphologic changes by WHO and unmodified RECIST criteria alone may not be sufficient to estimate tumor response in patients who received new targeted therapies [6][7][8].
There is an increased need for reliable imaging methods associated with early tumor response considering changes in other tumor characteristics such as tumor viability, metabolic activity, and attenuation. Several studies have reported low reliability of RECIST in evaluating treatment response in different tumors, such as hepatocellular carcinoma (HCC) [9], prostate cancer [10], brain glioblastoma [11], bone metastasis [12], malignant pleural mesothelioma (MPM) [13], and disseminated or ill-defined lymphoma [14]. Therefore, various specific criteria other than RECIST for response assessment of different tumors have been proposed (Table 2) [15]. According to RECIST, the selection of targeted lesions may be different among the readers for patients with multiple lesions. Variability may include scanrescan and both intra-and inter-observer inconsistency between two repeat readings of the same scan. Tumor size changes between interval studies may arise from true tumor changes and concomitant errors in measurement.
In recent years, several advanced imaging techniques are available to quantitatively assess tumor status and predict treatment response. The new medical imaging biomarkers change radiology culture towards more quantification and standardization. Beyond RECIST, images including dynamic contrast-enhanced (DCE) computed tomography (CT) or magnetic resonance imaging (MRI), diffusion-weight MR imaging (DWI), magnetic resonance spectroscopy (MRS), 18 F-fluorodeoxyglucose (FDG) positron emission tomography (PET) are emerging as promising imaging techniques in the assessment of tumor response to the new anti-cancer drugs [29]. This review evaluates the current RECIST with their limitations and the new emerging concepts of imaging biomarkers in oncology.

Anatomic or morphological approaches WHO criteria
The WHO established the first standardized approach to evaluate treatment responses of solid tumors based on   [16] modified RECIST (mRECIST) • To resolve limitations of anatomic tumor response metrics when applied RECIST 1.1 to moleculartargeted therapies or locoregional therapies in HCC. • Reassessment of progression that could be misinterpreted in RECIST 1.1 due to the natural progression of chronic liver disease (ascites, enlargement of lymph nodes, etc.). • Only well-delineated, arterially enhancing lesions can be selected as target lesions.
• Number of target lesions: up to 5 lesions (≤ 2 in any one organ).
• Short axis of porta hepatis lymph nodes ≥ 20 mm or other lymph nodes ≥ 15 mm are considered as malignant. imaging studies in 1979 [30]. However, the WHO criteria (  [3]. The original RECIST version 1.0, provided definitions for "measurable lesion" and "non-measurable lesion" (Table 1). Measurable lesions must have the longest diameter of ≥ 10 mm on CT with a slice thickness of ≤ 5 mm or the longest diameter of ≥ 20 mm on chest radiography [3]. Non-measurable lesions include other lesions that do not meet the criteria as measurable lesions, such as small lesions with a longest diameter less than 10 mm, bone metastases without a soft-tissue component, pleural tumor seeding, lymphangitic tumor spread, and peritoneal or leptomeningeal tumor diseases. After identifying measurable and non-measurable lesions, the RECIST includes the terms "target" and "non-target" lesions. Target lesions are selected based on their size (lesions with the longest diameters) and suitability/conspicuity for accurate repeated measurements. Target lesions include all measurable lesions (up to 5 per organ and 10 in total) that are recorded and measured at baseline. All other lesions were identified as non-target lesions. As several questions emerged, a revised RECIST guideline (RECIST 1.1) was developed in 2009 (Table 1) [4]. Major changes in RECI ST 1.1 include the number of target lesions, assessment of pathologic lymph nodes, redefinition of disease progression, clarification of unequivocal progression of nontarget lesions, and inclusion of 18 F-FDG PET in the detection of new lesions [4]. In RECIST 1.1, the maximum number of target lesions per organ was reduced from 5 to 2, and 10 to 5 in total. Besides, lymph nodes with a short axis ≥ 15 mm were considered measurable, pathological, and assessable as target lesions. Lymph nodes with a short axis < 10 mm were considered normal.
Lymph nodes with a short-axis between 10 and 15 mm were identified as non-measurable, non-target lesions.
Major changes from the WHO criteria to RECIST 1.1 are summarized in Table 1.
Evaluation of overall response in RECIST 1.1 is based on tumor responses in target and nontarget lesions (Table 3). For example, if there is any new lesion, the overall response is always PD. In the absence of a new lesion, complete response (CR) is the complete disappearance of all target and nontarget lesions. However, several studies have reported the low reliability of RECI ST 1.1 in evaluating treatment response in certain tumors. In HCC, a viable tumor is defined as showing intratumoral arterial enhancement on dynamic CT or MRI. Therefore, measurement of tumor enhancement is used as a surrogate biomarker of a viable tumor (Fig. 1), and tumor necrosis induced by treatment is considered as a response assessment [9]. Several specific criteria other than RECIST for HCC have thus been developed. The criteria include modified RECIST [16] (Table 2), European Association for the Study of the Liver (EASL) criteria [9], and the Response Evaluation Criteria in Cancer of the Liver (RECICL) criteria [32]. The RECIST is also not effective in evaluating treatment response for prostate cancer as there is not enough objective and meaningful measurement of disease progression [10]. In neuro-oncology, RECIST has limited use as the unidimensional measurements do not accurately measure the irregular or asymmetric margins of glioblastoma; besides, it did not take the use of steroids or clinical status into account [11]. The Response Assessment in Neuro-Oncology (RANO) criteria (Table 4) based on McDonald criteria in 2010 contemplated on non-enhancing tumor lesion on T1/T2-weighted imaging and fluid-attenuated inversion recovery (FLAIR) MRI. For bone metastasis, RECIST is limited to measurable metastases or unequivocal progression of the unmeasurable disease. For an accurate response assessment of bone metastases requires visualizing the tumor size as well the structural and metabolic changes in the bone (Fig. 2). This was addressed by the MD Anderson (MDA) bone response criteria (supplemental file) which updated the WHO criteria by expanding radiographic assessment and incorporating both CT and MRI [12,19]. In MPM, the measurement of the longest unidimensional diameter in  [28] • Unidimensional (longest diameter).
• New lesions are recorded separately, not included in the sum of lesions for target lesions identified at baseline. • Defining unconfirmed progressive disease (iUPD) and confirmed progressive disease (iCPD).
• iCPD: if additional new lesions appear or an increase in size of new lesions (≥ 5 mm for sum of new target lesion or any increase in new non-target lesion) on next cross-sectional imaging after iUPD.
tumor mass along the curved chest wall before and after the treatment is difficult that resulted in the development of modified RECIST. The modified RECIST for MPM considers tumor thickness perpendicular to fixed structures such as the chest wall or mediastinum in two positions on the same transverse slice of CT scan, and a sum of six measured values on three different levels was used for evaluation ( Fig. 3) [13]. In 1999, an international working group of clinicians, radiologists, and pathologists published Cheson criteria for response assessment and outcomes measurement in lymphoma [33]. Cheson criteria were adopted widely by clinicians and were used in the approval process for several new agents. However, the Cheson criteria were revised in 2007 because of identified limitations and the increased use of 18 F-FDG PET, immunohistochemistry, and flow cytometry in lymphoma. For lymphoma, the current standard response criteria are the revised Cheson criteria [20] (Table 5) and Lugano criteria [21] (supplemental file) based on PET or bidimensional tumor measurements on CT for non-FDG avid lesions [21]. In revised Cheson criteria, target lesions are defined based on the different organs (lymph nodes, liver or spleen, other organs) by both CT and PET scans, as well as the clinical examination and bone marrow biopsy. The Lugano classification, published in 2014, eliminated ambiguity and improved evaluations in lymphoma. It included standardized staging criteria for FDG-avid lymphomas using a five-point (Deauville) scale, defined splenomegaly as > 13 cm based on CT image, removed requirements for bone marrow biopsy for routine staging in Hodgkin's lymphoma and most diffuse B-cell lymphoma, and revised the definitions for PD. In recent years, new cancer immunotherapies such as immune checkpoint inhibitors, especially for advanced nonsmall cell lung cancer (NSCLC), have drawn greater attention [34]. Due to their peculiar mechanism, immunotherapies can determine unusual response patterns on imaging that cannot be correctly evaluated with the traditional RECI ST. Several immune-related response criteria including irRC, irRECIST, and iRECIST were proposed and applied in clinical trials for immunotherapies ( Table 2). One of the most common and challenging condition for the morphological evaluation of the response is the "pseudo-progression", a condition where the target lesion continues to grow at the first imaging study and then remains stable, shrinks in size, or disappears during the subsequent imaging follow-up (Fig. 4). The proposed hypotheses for these morphologic changes include (i) persistent tumor growth during the  Pseudo-progression has been reported for anti-programmed cell death protein-1 (anti-PD-1), anti-programmed deathligand 1 (anti-PD-L1), and anti-cytotoxic T lymphocyteassociated antigen 4 (anti-CTLA-4) agents not only in lung cancer but also other cancers, including melanoma, renal cell carcinoma, and bladder cancer [35][36][37].

Special considerations in lung cancer
Conventionally, lung cancer size is generally measured on lung window imaging and includes both ground-glass opacity (GGO) and solid components. The size of GGO within lung cancer generally does not vary markedly even after effective chemotherapy. Therefore, size change in the solid component of part-solid lung cancer may be a more accurate reflection of the actual tumor response to anti-cancer chemotherapy. Besides, intratumoral cavitation and necrosis caused by antiangiogenesis may also indicate tumor response (Figs. 5 and 6). Lee et al. [38] proposed a CT response criterion based on consideration of tumor components (GGO and solid part) and the presence of cavitation, necrosis, and attenuation changes for response assessment in NSCLC patients who underwent epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor (TKI) thertabapy (Figs. 5 and 6).

Special considerations in bone metastasis
Bone is one of the most common metastatic sites in lung cancer, breast cancer, or prostate cancer [39,40].  Treatments for bone metastasis include chemotherapy and endocrine therapy. The treatment response is evaluated by imaging modalities such as radiography, skeletal scintigraphy, CT and MRI, and PET. Accurate response assessment of bone metastases requires evaluation of both the tumor burden and the osteolytic or osteoblastic changes in the bone lesion (Fig. 2). As the RECIST focuses predominantly on the physical measurement of solid tumors, bone metastasis is not easily measured with a ruler and is designated as unmeasurable. Patients with only bone metastases after resection of a primary tumor are not eligible for clinical trials due to no measurable disease. Therefore, the absence of measurable tumors can significantly influence treatments in these patients. The MDA criteria (supplemental file) incorporates information from CT imaging into that of the WHO criteria, specifically for the response assessment in bone metastases. It divides tumor response into 4 standard categories and includes quantitative and qualitative evaluation of the bone metastases. Hamaoka et al. [19] reported that the MDA criteria are superior to the WHO criteria in predicting progression-free survival in breast cancer patients with bone metastasis. In comparison to RECIST 1.1, the MDA criteria allow more metastatic bone lesions to be considered as a measurable disease by allowing physical measurement of well-defined bone lesions regardless of soft tissue extension, allowing subjective assessment of illdefined lesions, and considering healing sclerosis. Other functional imaging criteria such as Positron Emission Tomography Response Criteria in Solid Tumors (PERC IST) ( Table 6) allow bone metastases to be measured by assessing tumor metabolism in the absence of anatomic change [12].

Clinical limitations and challenges of morphological evaluation
According to RECIST, the standard method to assess the response of solid tumors to chemotherapy is to decide those target lesions first and perform a uni-   Variability can be caused by scan-rescan variability and both intra-and inter-observer variability between two repeat readings of the same scan. Oxnard et al. [42] reported that changes of tumor size less than 10 % can be a result of the inherent variability in patients with advanced NSCLC. This variability was the greatest in the measurement of small tumors and the significant importance for the accurate determination of disease progression.
The measurement of the entire tumor volume overcomes some of the limitations, as to improve the ability to reliably detect small changes in measurements, and to increase statistical power per subject in trials [43]. Neither the RECIST nor the WHO criteria include volume measurement because of limitations in past diagnostic imaging techniques and the available measurement methods. But with the advent of thin-section CT and commercially available tumor segmentation software, it is now possible to obtain image data sets with spatial resolutions adequate to measure tumor volumes [44]. Zhao et al. [45] suggested that measuring volumetric changes in tumor dimension may hold the potential to be an earlier biomarker of tumor regression or progression. The changes in tumor volume may be assessed as early as 3 weeks after the initiation of gefitinib (Iressa) treatment, whereas a lower magnitude of changes in unidimensional and bidimensional measurements was seen during the same period [45].

Functional, metabolic, and other non-morphological approaches
Functional and metabolic imaging techniques can integrate pathological, physiological, and morphological changes, and serve as potential early predictors for therapeutic response [24,46]. The microscopic changes in the tumor microenvironment are detected and early evaluation of response to therapy such as tumor attenuation/enhancement, perfusion, oxygenation, metabolism, etc., are possible.

Evaluation of tumor attenuation/enhancement
Tumor enhancement/attenuation changes in CT are important imaging biomarkers for the assessment of tumor responses [22,47,48]. In patients receiving molecular targeting agents, a decrease of tumor attenuation on CT indicates a response to therapy, even in the absence of decreased tumor size as defined by RECIST. In contrast, tumor progression may demonstrate patterns of new intratumoral enhancing lesions rather than tumor size increase. Choi criteria [22] (Table 7) were the first to introduce this CT parameter for evaluating the effect of imatinib mesylate (Gleevec), a TKI, in gastrointestinal stromal tumors (GIST). For malignant GISTs, early dramatic changes in the extent of tumor attenuation are often observed after using imatinib (Fig. 7) [49]. However, morphological changes in unmodified RECIST alone may not be sufficient to estimate the effect of imatinib in GISTs, especially at an early stage of treatment that are poor predictors of clinical benefit. Choi criteria define a good response by a 10 % decrease in tumor size or a 15 % decrease in CT attenuation (Table 7). Choi criteria also define PD by the (1) appearance of new lesions, (2) appearance or increase in the size of intratumoral nodules, or (3) tumor size increase by more than 20 % without post-treatment hypoattenuation change.
Similar phenomena were observed in other solid tumors, including HCC, colorectal liver metastases, renal cell carcinomas (RCC), and sarcomas [22,47,48]. In 2000, the EASL agreed that estimating the reduction in viable tumor volume (recognized as poorly-enhanced areas on dynamic CT or MRI) should be considered for locoregional therapies such as trans-arterial chemoembolization (TACE) or radiofrequency ablation (RFA) in HCC [9]. In HCC, several studies showed a poor correlation between the clinical benefits and conventional response criteria in treatments by sorafenib and locoregional therapies such as TACE or RFA [50,51]. Therefore, a modified RECIST (mRECIST) was described by Lencioni et al. [16] in 2010. The mRECIST for HCC was developed based on the concept that a viable tumor is defined as showing intratumoral arterial enhancement by contrast agent during dynamic CT or MRI (Fig. 1) [9]. Therefore, measurement of tumor enhancement can be used as a surrogate biomarker of a viable tumor. Furthermore, tumor necrosis induced by treatments should be considered as a response assessment. Besides, for patients with colorectal liver metastasis, tumor attenuation correlates well with survival rate after bevacizumab treatment, whereas RECIST criteria do not correlate with patient survival [52]. For metastatic RCC treated with sunitinib, the Choi criteria also had a better predictive value than RECIST in defining patients who benefit from therapy [53].

CT and MRI perfusion imaging
The basis for using CT and MRI perfusion in oncology is that the microvascular changes of angiogenesis are reflected by neovascularization in tumor cells [54]. These techniques can quantify regional tumor blood flow (BF), tumor blood volume (BV), micro-vessel permeability, contrast medium extraction fraction, and extraction fraction on plasma and interstitial volumes. Clinical applications of CT and MRI perfusion include lesion characterization, tumor staging, predicting therapeutic response and clinical outcome, and evaluation of tumor relapse and drug resistance [55].
The degree of CT perfusion is related to tumor aggressiveness in many types of cancer, and highly vascularized tumors are associated with a poor prognosis [54,56,57]. Higher baseline BF on CT perfusion predicts better response to chemotherapy and radiotherapy in different tumors. Bellomi et al. [58] reported that patients with   [59] showed that lower CT perfusion values had a significantly higher local failure rate. Hypervascular tumors allow delivery of chemotherapy and may have more oxygenation for greater radiosensitivity [60,61]. Different antiangiogenic agents such as antibodies targeting vascular endothelial growth factor (VEGF) and small-molecule TKIs have been developed [57,62]. Higher pretreatment CT perfusion values are associated with a better response rate to antiangiogenic drugs [63,64]. As compared to RECIST and tumor attenuation, Jiang et al. [65] proposed that CT perfusion is a more sensitive imaging biomarker in advanced HCC patients treated with a combination of anti-angiogenic and conventional chemotherapies. MRI perfusion also has the ability to analyze underlying tumor angiogenesis by parameters related to tumor perfusion and permeability after the intravenous injection of gadolinium-based contrast agents [66]. Depending on the technique utilized, MRI perfusion can provide information about tissue cerebral blood volume and perfusion (using dynamic susceptibility contrastenhanced [DSC] MRI), or micro-vessel permeability and the extracellular space (using T1-weighted dynamic contrast-enhanced [DCE] MRI). DSC-MRI is usually used for the evaluation of brain tumors, and the most commonly calculated parameters are relative cerebral blood volume (rCBV), relative cerebral blood flow (rCBF), and mean transit time (MTT). Pharmacokinetic analysis of DCE-MRI is the most widely used method for the quantitative measurement of vessel permeability changes (Fig. 8). With the Tofts model [67], an evaluation of the injected contrast agent leaking into the extravascular-extracellular space, and tissue perfusion and permeability becomes possible. The volume transfer constant K trans (wash-in rate; unit: min − 1 ) describes the forward leakage rate of the contrast medium. For blood vessels where the leakage is rapid, perfusion determines the contrast agent distribution and K trans is similar to tissue blood flow per unit volume [68]. This is typically reported in breast tumors where the endothelium is leaky and other high permeability situations and is termed a flow-limited situation. For the prediction of therapeutic  Morabito et al. [69] reported that both blood perfusion (rCBV) and micro-vessel permeability (K trans ) are useful tools for differentiating tumor recurrence from radiation necrosis in brain tumors. Quantitative assessment of contrast enhancement kinetic curve on breast DCE-MRI resulted in excellent diagnostic performance for differentiation of malignancy (Fig. 8) [70]. Wedam et al. [71] reported that DCE-MRI parameters including K trans , kep (reverse rate constant from tumor to the vascular space), and ve (extravascular volume fraction) could be used as an early imaging biomarker for monitoring treatment response in breast cancer patients receiving bevacizumab.
Other cancers that can be monitored by DCE-MRI include radiotherapy in rectal cancer [72], metastatic renal cell carcinoma treated with sorafenib [73], androgen withdrawal therapy in prostate cancer [74], and radiofrequency ablation in HCC [75]. It is known that successful treatment usually leads to decreases in BF and permeability in assessing response to chemotherapy, radiotherapy, and anti-angiogenic therapy [76].

Diffusion-weighted MR imaging (DWI)
The consensus concerning DWI as a cancer imaging biomarker was reached in 2008 at the International Society for Magnetic Resonance in Medicine Meeting [77]. Therefore, DWI has been widely used as an imaging biomarker in the characterization of malignancy, determination of lesion aggressiveness, and monitoring response to a variety of treatments [77][78][79]. In theory, DWI provides biomedical information related to tissues and structures of interest-based on the measurement of thermally induced "Brownian motion" of water molecules shown by apparent diffusion coefficient (ADC) value [80]. The information provided by DWI includes tissue perfusion, cellularity, extracellular space distribution, and the integrity of cell membranes. Thus, unusual imaging findings on DWI could be an early predictor of biophysiological abnormality [77]. In oncologic imaging, DWI has been linked to tumor aggressiveness and treatment response [77]. Parameters derived from DWI are appealing as imaging biomarkers because the acquisition is rapid and noninvasive, requiring no exogenous contrast agents or exposure to ionizing radiation. In oncologic practice, the measurements of water diffusivity are used as biomarkers of tissue properties. Since tissue water movements are not ''free'' but impeded by cells, extracellular matrix, and other molecules, the measurement of tissue water diffusivity is often termed the ADC value. The biomarkers used in oncology include relative signal intensity at different b-values, water diffusivity (D), perfusion fraction (F p ), ADC total , fractionated ADC (ADC fast and ADC slow ), and fractional anisotropy (FA) [81]. F p represents the contribution of blood microcirculation and microscopic flow to signal decay. ADC fast is calculated using low b-values (0-100 s/mm 2 ) and is dominated by this perfusion component of the total tissue diffusivity. At higher b-values (> 100 s/mm 2 ), the perfusion component is largely extinguished, so ADC slow measurements are more heavily determined by water diffusion within the cellular matrix. For most clinical studies, only ADC total is reported, which is the total ADC including perfusion contributions. In the above discussion, it is assumed that water movements in tissues are nondirectional (isotropic). However, in some tissue such as brain white matter and renal tubules, water mobility in normal tissues can be directional (anisotropic). A commonly used measurement of anisotropy is FA, which is a dimensionless quantity with range between 0 (nondirectional, isotropic) to 1 (highly directional, anisotropic) [82].
The DWI offers useful clinical information at all stages of cancer, including detection, diagnosis, staging, monitoring therapy response, assessing recurrence, and developing pharmaceutical drugs [77]. The applications of DWI as an early biomarker for the prediction of treatment outcomes in different tumors such as brain glioblastoma, hepatic tumor, lung cancer, and primary bone sarcomas had been reported ( Fig. 9) [83][84][85][86][87]. Pretreatment low ADC values correlate with poor survival in malignant brain astrocytomas independent of tumor grade (Fig. 9) [88]. Higher pretreatment ADC values were associated with better therapeutic response in vascular disruptive agents [89]. In contrast, several clinical studies have noted that higher pretreatment ADC values respond less favorably to treatments [90][91][92], reflecting the presence of microscopic and macroscopic necrosis, which is recognized to be associated with poorer therapeutic outcomes [93]. Association between necrosis and hypoxia in tumors is probably involved. Hypoxia in tumor cells mediates resistance to chemotherapy, radiation, and photodynamic therapy [93]. The association between higher pretreatment ADC values and the poorer response had been reported in rectal cancer, liver metastasis, and breast cancers [94][95][96]. Furthermore, successful treatment is reflected by increased ADC values in several different tumors, including breast cancer, HCC, malignant brain tumors, and primary bone sarcomas [97][98][99][100].

Magnetic Resonance Spectroscopy (MRS)
The alterations in metabolite levels are best detected both noninvasively and quantitatively by MRS. Although MRS is usually applied in brain lesions, it can also be used for detection, staging, evaluation of aggressiveness, and assessment of therapeutic response in the breast, prostate, hepatic, and other cancers [46]. MRS can be procured from several nuclei in the body, such as 1-Hydrogen (1 H), 31-Phosphorus, and 19-Fluorine. Typically, 1 H-MRS is the most common clinically used method because of the high sensitivity to this nucleus, easy accessibility, and the abundant existence of hydrogen in metabolites. MRS procedure begins with the acquisition of MR images and then assesses the "metabolite spectrum" in the region of interest. In single voxel spectroscopy (SVS), a single voxel (volume of tissue) is located in the tumor or lesion where the metabolism may be damaged as a result of patient disease. Another technique of MRS is known as chemical shift imaging (CSI) [101]. In CSI, a large volume divided into several smaller voxels is selected to produce all voxel's spectrum simultaneously. Usually, SVS is the preferred choice when accurate quantification is required, and CSI is used to provide an overall vision of spatial distributions of the metabolite [102,103]. In brain tissue, N-acetyl aspartate (NAA), creatine (Cr), and choline (Cho) are the most important metabolites as an MRS signal [104]. The derivatives of Cr form an important system for energy metabolism, and their intensity is considered fixed. Therefore, they are applied in the computing of metabolite ratios. Brain tumors commonly cause an increase in Cho concentration. Cho level is related to the ability of proliferation and existence of malignancy, and an increased Cho signal is a marker of the presence of a brain tumor [105]. In contrast to Cho, the signal of NAA, which is a neuronal marker, typically decreases in brain tumors. Since the level of Cho and NAA signal changes in brain tumors, the calculation of this Cho-NAA ratio is helpful in the interpretation of brain tumor's MRSI spectra [106]. The anaerobic metabolism in tumor cells also causes the appearance of lactate peak in the brain tumor's spectra. The increase in lactate and lipid peaks are the markers of tumor progression and transformation from a lowgrade to a high-grade tumor [107].
For breast cancer, Roebuck et al. [108] first proposed that the Cho peak can be applied as a sign of malignancy (Fig. 10). Several studies also showed that Cho peak is a useful marker in malignant breast lesions, and it is not visible in normal tissues or benign tumors [109,110]. Further, Jagannathan et al. [111] first reported the Cho peak is diminished in 89 % of cases that underwent chemotherapy. These promising results suggest that MRS increases the sensitivity, specificity, and accuracy of breast MRI and will be a useful tool in the diagnosis and management of breast cancer [112]. In prostate cancer, CSI of MRS has shown that it has high diagnostic accuracy [113]. The MRS gathers metabolic information of prostatic tissue by calculating the relative concentrations of citrate, Cr, and Cho. Further, prostate adenocarcinoma can be differentiated from normal adjacent tissue based on the (Cho + Cr)/citrate ratio with a cut-off value of 0.8 [114,115]. H1-MRS can also be applied in the diagnosis of gastrointestinal tumors and malignant ovarian tumors [116,117]. These results revealed that most cancer lesions can be determined by the increase of Cho and lactate peaks with a decrease of lipid peak [118].

Metabolic imaging
For newer anti-cancer therapies that stabilize disease rather than reduce tumor size at the beginning of treatments, the 18 F-FDG PET offers particularly valuable information in these cases. PET can assess tissue metabolism by using radiolabeled molecules, most commonly 18 F-FDG, a glucose analog. 18 F-FDG PET-CT or MRI is of value in the differentiation of benign and malignant tissues, preoperative staging, recurrent disease detection, and identification of early tumor response to therapy. Wahl et al. [24] proposed guidelines for the standardization of response criteria for FDG PET, the so-called PERCIST criteria ( Table 6). The standardized uptake value (SUV) normalized by lean body mass (SUL) represents a quantitative assessment of uptake in a tumor region of interest. The SUV/SUL is based on a ratio between tracer uptake within a tumor and homogeneous distribution of tracer within the patient body. By reflecting the change in tumor metabolism, 18 F-FDG PET establishes a method to measure tumor response in the absence of marked anatomic changes (Fig. 11) [119]. A decrease in FDG uptake indicates treatment response or improves survival rates in patients with solid tumors such as breast cancer [120], esophageal cancer [121], lung cancer [122,123], osteosarcoma [124], and other tumors [125]. 18 F-FDG PET provides more rapid response data than morphological measurements [126,127]. In patients with NSCLC, 18 F-FDG PET is a staging tool, and the SUV provides prognosis before and after the treatment [122,123]. It prevented futile thoracotomies in lung cancer [128] and stratified patients into surgical versus palliative groups in colorectal cancer [129]. However, the SUV changes are also influenced by other factors such as vascular delivery. In primary breast tumors, a reliable dropin SUVs indicating a tumor response is seen only for patients with high initial SUVs [130]. Therefore, a reduction in PET metabolism caused by chemotherapy may be dependent on pre-therapy vascular delivery [131].

Radiomics approach
Radiomics is a possible exciting complement to RECIST for monitoring and predicting therapeutic response. Both CT and MRI radiomics analysis are new field in medical imaging. This approach is based on computerized extraction of several quantitative imaging features, and uses these data for the medical decision, prediction, and monitoring response to therapy [132]. A general pipeline of radiomics analysis includes feature extraction, feature selection, and prediction. Feature extraction quantitatively defines the imaging parameters from the specified areas of the images. Feature selection evaluates the feature importance based on the objectives, and the prediction model is established by selected features [132]. Radiomics extracts a large number of quantitative imaging features from a medical image and then analyses these features by a series of machine learning algorithms. The extracted imaging features are related to the underlying anatomical microstructure and biophysical processes such as genetic expression, tumor proliferation, and tumor neovascularization [133]. Radiomics in texture and shape analysis had been widely used to evaluate medical images with promising results [134,135]. Recently, radiomics analysis is emerging as a comprehensive quantitative method to diagnose and evaluate response to therapy in brain tumors, head and neck cancer, breast cancer, liver tumor, prostate cancer, rectal cancer, non-small-cell lung cancer, and metastatic RCC [136][137][138][139][140][141][142][143]. With radiomics, texture analysis reveals visually imperceptible information that extends beyond radiology to histopathology and provides predictors for diagnosis, prognosis, and therapeutic planning for cancer patients [136][137][138][139].
Radiomics is especially a possible useful tool to predict response to radiation therapy, different chemotherapies, and immune checkpoint inhibition treatments, especially in lung cancer [144,145]. It can be potentially integrated into the normal clinical workflow to identify lung cancer patients who would benefit most from therapy [144]. Huynh et al. [145] reported that CT-based radiomic shape and tumor heterogeneity features could predict the treatment response to stereotactic body radiation therapy in early-stage NSCLC. In locally advanced NSCL C, Rakshit et al. [146] showed that certain textural features are radiomics predictors for response to pemetrexed chemotherapy, with an AUC of 0.81. Velcheti et al. [147] developed a pretherapy CT radiomics-based predictive model for the prediction of tumor response to Nivolumab in locally advanced NSCLC, with an AUC of 0.84. Other studies applying radiomics on NSCLC also showed very positive and promising results [148,149].

Conclusions
Conventionally, tumor response is evaluated basically and readily by the use of RECIST 1.1. However, the criteria mainly focus on tumor dimensional changes and do not reflect other functional, metabolic, and nonmorphological changes that may occur in moleculartargeted therapy, immunotherapy, and local treatments. Since morphological changes in unmodified RECIST 1.1 alone may not be sufficient to estimate the treatment response in cancer patients, several modified criteria have been proposed to improve the response assessment in different specific tumors. Due to the rapid progress in anti-cancer therapies nowadays, the new imaging techniques such as CT/MRI-based perfusion, DWI, MRS, PET, and radiomics should be recognized by radiologists and oncologists for tumor response evaluation. On the other hand, although imaging is utilized for evaluating tumor response in daily oncological practice, the situation in which imaging is used differs to some degree between clinical trials and routine clinical practice [150]. The challenges of standardizing imaging in the multicenter clinical trials include the processes of image acquisition, data analysis, and radiological review [150]. The emergence of new treatment paradigms and the trend toward personalized treatment should be accompanied by the evolution of response assessment. Although these new imaging techniques make clinically effective contributions to cancer care as a decisionmaking method for personalized medicine [151], it may be cost prohibitive and challenging to implement in some institutions at this moment. To resolve this problem, the aim toward developing more affordable and effective alternatives is necessary. In our opinion, integrating multiparametric quantitative information from different imaging modalities for precise personalized medicine and evaluation is the goal of cancer treatment in the future