**2. Endoscopic assessment of polyp histology**

The key factor in adopting the "resect and discard" strategy is the endoscopic evaluation of polyp histology, since this information is necessary to plan the next surveillance interval. Moreover, the presence of suspicious endoscopic features may prompt a polyp to be submit‐ ted to pathologic assessment. Therefore, a reliable endoscopic method of evaluating histolo‐ gy is needed.

In recent years several imaging-enhancing technologies have emerged as an adjuvant for di‐ agnosing and evaluating colorectal lesions [34]. High-resolution and magnification endo‐ scopes allow enlarging the image and discriminating details. These endoscopes are often used in combination with chromoendoscopy, which involves the topical application of dyes at the time of endoscopy to enhance tissue characterization. Narrow-band imaging (NBI) is a technology that applies narrow-bandwidth filters to white light endoscopy allowing dis‐ crimination of mucosal vascular net. Fuji Intelligent Color Enhancement (FICE) and i-Scan are based on the same physical principles as NBI but are not depending on optical filters but on a postprocessing image system. All these technologies have been evaluated in the predic‐ tion of histology of colon polyps.

#### **2.1. High-resolution/magnification endoscopy and chromoendoscopy**

The usefulness of this technology in assessing histology is based on the pit-pattern classifica‐ tion proposed by Kudo which is intended to differentiate between non-neoplastic, neoplas‐ tic and malignant polyps. Following this classification patterns I and II correspond to nonneoplastic lesions and patterns III to V to neoplastic ones. Type V suggests malignant transformation [35].

Several large case series evaluate the utility of pit-pattern analysis to differentiate neoplastic from non-neoplastic lesions. Generally speaking, positive predictive values (PPV) for neo‐ plastic lesion range between 70 to 100% and negative predictive values (NPV) between 70 and 99%. Studies with the largest number of lesions show an overall accuracy of 80-95% [36-38]. One study focused in diminutive lesions, reported an overall accuracy of 95% [39]. There are also some randomized controlled trials comparing magnification plus chromoen‐ doscopy to conventional chromoendoscopy. Konishi et al. [40] showed an accuracy of mag‐ nification colonoscopy in distinguishing non-neoplastic from neoplastic lesions < 10 mm in size of 92% vs 68% for conventional chromoendoscopy. Emura el al. [41] using a similar de‐ sign showing an overall accuracy of 95% vs 84%. These figures were similar whenthe sub‐ group of lesions ≤ 5 mm was analyzed. Conventional colonoscopy, chromoendoscopy and magnification chromoendoscopy were compared in the study by Fu et al. [42], and the latter was found to have the highest accuracy (95.6%).

Magnification chromoendoscopy has also been evaluated in the prediction of malignant his‐ tology and invasive depth of cancer with variable results. Overall, it seems that its sensitivi‐ ty and accuracy are lower. For instance, Bianco et al. [43] showed that endoscopic differentiation between invasive and noninvasive neoplasm had a PPV of 79% and a NPV of 95%. Hurlstone et al. reported an accuracy of 78% and a specificity of 50% [44]. Some au‐ thors use a modification of the Kudo classification with different subtypes of the type V pat‐ tern that may be quite cumbersome to use [45].

In conclusion, high-magnification chromoendoscopy allows the prediction of histology even in small and diminutive lesions, but is better differentiating nonneoplastic from neoplastic lesions than differentiating invasive from noninvasive neoplasms. Moreover, it must be kept in mind that overall accuracy is not 100%, despite the fact that a technology with a NPV of 95% for adenomatous histology fulfils the PIVI criteria for leaving suspected rectosigmoid hyperplastic polyps ≤ 5 mm in size in place [33].

#### **2.2. Narrow-band imaging**

resolution photograph and the polyp then would be resected and discarded. The endoscopic assessment of histology would be used to make an immediate recommendation regarding the next colonoscopy surveillance interval. Finally, when multiple diminutive rectosigmoid hyperplastic polyps are suspected endoscopically, histology can be established by real-time endoscopic assessment and documented by photography without the need of resection and

The key factor in adopting the "resect and discard" strategy is the endoscopic evaluation of polyp histology, since this information is necessary to plan the next surveillance interval. Moreover, the presence of suspicious endoscopic features may prompt a polyp to be submit‐ ted to pathologic assessment. Therefore, a reliable endoscopic method of evaluating histolo‐

In recent years several imaging-enhancing technologies have emerged as an adjuvant for di‐ agnosing and evaluating colorectal lesions [34]. High-resolution and magnification endo‐ scopes allow enlarging the image and discriminating details. These endoscopes are often used in combination with chromoendoscopy, which involves the topical application of dyes at the time of endoscopy to enhance tissue characterization. Narrow-band imaging (NBI) is a technology that applies narrow-bandwidth filters to white light endoscopy allowing dis‐ crimination of mucosal vascular net. Fuji Intelligent Color Enhancement (FICE) and i-Scan are based on the same physical principles as NBI but are not depending on optical filters but on a postprocessing image system. All these technologies have been evaluated in the predic‐

The usefulness of this technology in assessing histology is based on the pit-pattern classifica‐ tion proposed by Kudo which is intended to differentiate between non-neoplastic, neoplas‐ tic and malignant polyps. Following this classification patterns I and II correspond to nonneoplastic lesions and patterns III to V to neoplastic ones. Type V suggests malignant

Several large case series evaluate the utility of pit-pattern analysis to differentiate neoplastic from non-neoplastic lesions. Generally speaking, positive predictive values (PPV) for neo‐ plastic lesion range between 70 to 100% and negative predictive values (NPV) between 70 and 99%. Studies with the largest number of lesions show an overall accuracy of 80-95% [36-38]. One study focused in diminutive lesions, reported an overall accuracy of 95% [39]. There are also some randomized controlled trials comparing magnification plus chromoen‐ doscopy to conventional chromoendoscopy. Konishi et al. [40] showed an accuracy of mag‐ nification colonoscopy in distinguishing non-neoplastic from neoplastic lesions < 10 mm in size of 92% vs 68% for conventional chromoendoscopy. Emura el al. [41] using a similar de‐ sign showing an overall accuracy of 95% vs 84%. These figures were similar whenthe sub‐

**2.1. High-resolution/magnification endoscopy and chromoendoscopy**

pathological evaluation [33].

tion of histology of colon polyps.

transformation [35].

gy is needed.

**2. Endoscopic assessment of polyp histology**

180 Colonoscopy and Colorectal Cancer Screening - Future Directions

#### *2.2.1. Predicting histology by means of vascular features*

Angiogenesis is a main step in the progression of neoplasms; therefore the diagnosis based on vascular morphological changes seems ideal for early detection and diagnosis of co‐ lon neoplasms. NBI enhances the visibility of the capillary network on the surface layer of the mucosa.

Normal mucosa displays a regular hexagonal or honeycomb-like pattern of capillary vessels around the crypt of the gland. This capillary meshwork, named meshed capillary (MC), is invisible or faintly visible (Figure 2a). In the neoplastic lesion, vessels grow thicker, with increasing diameter size, disruption and rise of vessel density as the lesion progresses. There‐ fore, recognizing the lesion becomes easier because it appears as a brownish area (Figure 2b).

**Figure 2.** NBI image of normal mucosa (a) and a diminutive adenoma (b)

Several studies have evaluated the performance of NBI in characterizing colorectal lesions, focusing in the characteristics of the vascular capillary network. Generally speaking, NBI sensitivity and specificity for diagnosing neoplastic lesions ranges between 77% and 99% and 59 – 100% respectively (table 2). This heterogeneity may be explained by the use of dif‐ ferent descriptions of vascular networks. Examples are, brown blob or dense vascular net‐ work to predict neoplasia [46-48]; fine capillary network, dark dots, light rounds, tubular or gyrus like [49]; microvessel thickness (invisible, thin, thick) and microvessel irregularity (in‐ visible, regular, mildly irregular, severely irregular) [50]; vascular patter intensity (weaker, the same or darker than the surrounding mucosa) [51]; fine vascular network or dilated corkscrew type vessels and abnormal branching patterns [52]; and finally, capillary pattern (CP type I: invisible or faintly visible, CP type II: capillaries elongated and thicker and CP type III: capillaries of irregular sizes, thicker and branched) [53-55].Other causes of heteroge‐ neity are the use of magnification or high-resolution endoscopes since the results with the latter are not as encouraging (see section 2.2.5) [46,49,56], and finally, better results are re‐ ported by experts.

**Author Mag Patients/**

Values for each observer are shown.

AC: diagnostic accuracy.

**Author Mag Patients/**

*2.2.3. Predicting submucosal invasion*

**Lesions**

**Sensitivity (%) Specificity (%) PPV**

In vivo Optical Diagnosis of Polyp Histology: Can We Omit Pathological Examination of Diminutive Polyps?

**Sensitivity (%) Specificity (%) PPV**

Chiu [46] a Yes/No 133/180 87-95 88-72 96-92 67-80 87-90 Sano [53] Yes 92/150 96 92 97 90 95 Hirata [50] Yes 163/189 99 90 99 90 98 Hirata [48] Yes 99/148 99 94 99 94 99 Rastogi [49] No 40/123 96 86 90 95 92 Kanao [55] Yes 223/289 95 100 100 20 99 Henry [54] No 42/126 93 88 909 91 91 Ignjatovic [56] Yes/No 48/80 93-74 59-56 - - 76-85

**Table 2.** Vascular pattern analysis with NBI for prediction of adenomatous histology. Mag: use of optical magnification; PPV: positive predictive value; NPV: negative predictive value; DA: diagnostic accuracy. a

Machida [57] Yes 34/43 100 75 91 100 93 East [51] a Yes 20/33 86-77 80-60 - - 84-72 Tischendorf [52] Yes 99/200 90 89 93 84 90 Van den Broek [59] Yes 100/208 90 70 69 90 78

observer are shown. Mag: use of optical magnification; PPV: positive predictive value; NPV: negative predictive value;

A systematic review which included 6 reports published until 2008 comparing NBI (pit pat‐ tern and vascular assessment) and chromoendoscopy showed a pooled sensitivity, specifici‐

NBI has also been evaluated to diagnose early colorectal neoplasia and submucosal inva‐ sion. Katagiri et al. [61] used the capillary pattern classification in colon adenomas. Those showing CP type III harbored HGD or invasive cancer. In a recent report this group further developed this classification expanding CP type III in group IIIA (visible microvascular ar‐ chitecture and high microvessel density with lack of uniformity, branching and curtailed ir‐ regularity) and group IIIB (nearly avascular or loose microvascular area). This detailed classification allowed differentiation between lesions with Sm1 submucosal invasion from Sm2-Sm3 with a sensitivity, specificity and diagnostic accuracy of 84.8%, 88.7% and 87.7% respectively [62]. Hirata et al.[50] found that the accuracy of diagnosis of submucosal mas‐

**Lesions**

**Table 3.** Pit pattern analysis with NBI for prediction of adenomatous histology a

ty and overall accuracy of 92%, 86% and 89% respectively [60].

**(%)**

**(%)**

Two observers. Values for each

**NPV (%)**

**NPV (%)**

http://dx.doi.org/10.5772/53205

**DA (%)** 183

Two observers.

**DA (%)**

#### *2.2.2. Predicting histology by means of pit pattern evaluation*

Most of the published studies, mainly from Japan, use optical magnification in combination with NBI, and the performance of pit pattern analysis with NBI ha salso been assessed (table 3). Sensitivity for neoplastic lesion ranges between 86 and 100%, while specificity ranges be‐ tween 84 and 100%. Some studies have compared NBI with chromoendoscopy showing sim‐ ilar diagnostic accuracy, suggesting that NBI could replace chromoendoscopy in the diagnostic evaluation of colon lesions [46, 47, 52]. However, the original pit pattern classifi‐ cation was not designed for NBI, and has not been validated for this purpose. NBI funda‐ mentals are different that those of chromoendoscopy. The latter uses dyes that lie inside the pits or stain their edges depending on the stain used while NBI highlights the capillary plexus that surrounds the opening of each pit. Machida et al. [57] described the use of NBI with magnification for pit pattern classification, showing that NBI was superior to conven‐ tional colonoscopy for pit pattern delineation but inferior to chromoendoscopy. The correla‐ tion between pit pattern analysis using chromoendoscopy and NBI is far from perfect especially for the pattern with the upmost clinical importance, type V. A study compared the pit pattern analysis obtained by NBI with stereoscopic examination and showed that the correlation was only 57% for type VN[58]. East et al. [51] found a kappa score of only 0.23 between both types of pit pattern evaluation. Better results were obtained by Hirata et al. [48] (78% of agreement for pit pattern VI and 100% for VN).


In vivo Optical Diagnosis of Polyp Histology: Can We Omit Pathological Examination of Diminutive Polyps? http://dx.doi.org/10.5772/53205 183


**Table 2.** Vascular pattern analysis with NBI for prediction of adenomatous histology. Mag: use of optical magnification; PPV: positive predictive value; NPV: negative predictive value; DA: diagnostic accuracy. a Two observers. Values for each observer are shown.


**Table 3.** Pit pattern analysis with NBI for prediction of adenomatous histology a Two observers. Values for each observer are shown. Mag: use of optical magnification; PPV: positive predictive value; NPV: negative predictive value; AC: diagnostic accuracy.

A systematic review which included 6 reports published until 2008 comparing NBI (pit pat‐ tern and vascular assessment) and chromoendoscopy showed a pooled sensitivity, specifici‐ ty and overall accuracy of 92%, 86% and 89% respectively [60].

#### *2.2.3. Predicting submucosal invasion*

Several studies have evaluated the performance of NBI in characterizing colorectal lesions, focusing in the characteristics of the vascular capillary network. Generally speaking, NBI sensitivity and specificity for diagnosing neoplastic lesions ranges between 77% and 99% and 59 – 100% respectively (table 2). This heterogeneity may be explained by the use of dif‐ ferent descriptions of vascular networks. Examples are, brown blob or dense vascular net‐ work to predict neoplasia [46-48]; fine capillary network, dark dots, light rounds, tubular or gyrus like [49]; microvessel thickness (invisible, thin, thick) and microvessel irregularity (in‐ visible, regular, mildly irregular, severely irregular) [50]; vascular patter intensity (weaker, the same or darker than the surrounding mucosa) [51]; fine vascular network or dilated corkscrew type vessels and abnormal branching patterns [52]; and finally, capillary pattern (CP type I: invisible or faintly visible, CP type II: capillaries elongated and thicker and CP type III: capillaries of irregular sizes, thicker and branched) [53-55].Other causes of heteroge‐ neity are the use of magnification or high-resolution endoscopes since the results with the latter are not as encouraging (see section 2.2.5) [46,49,56], and finally, better results are re‐

Most of the published studies, mainly from Japan, use optical magnification in combination with NBI, and the performance of pit pattern analysis with NBI ha salso been assessed (table 3). Sensitivity for neoplastic lesion ranges between 86 and 100%, while specificity ranges be‐ tween 84 and 100%. Some studies have compared NBI with chromoendoscopy showing sim‐ ilar diagnostic accuracy, suggesting that NBI could replace chromoendoscopy in the diagnostic evaluation of colon lesions [46, 47, 52]. However, the original pit pattern classifi‐ cation was not designed for NBI, and has not been validated for this purpose. NBI funda‐ mentals are different that those of chromoendoscopy. The latter uses dyes that lie inside the pits or stain their edges depending on the stain used while NBI highlights the capillary plexus that surrounds the opening of each pit. Machida et al. [57] described the use of NBI with magnification for pit pattern classification, showing that NBI was superior to conven‐ tional colonoscopy for pit pattern delineation but inferior to chromoendoscopy. The correla‐ tion between pit pattern analysis using chromoendoscopy and NBI is far from perfect especially for the pattern with the upmost clinical importance, type V. A study compared the pit pattern analysis obtained by NBI with stereoscopic examination and showed that the correlation was only 57% for type VN[58]. East et al. [51] found a kappa score of only 0.23 between both types of pit pattern evaluation. Better results were obtained by Hirata et al.

**Sensitivity (%) Specificity (%) PPV**

**(%)**

**NPV (%)**

**DA (%)**

ported by experts.

*2.2.2. Predicting histology by means of pit pattern evaluation*

182 Colonoscopy and Colorectal Cancer Screening - Future Directions

[48] (78% of agreement for pit pattern VI and 100% for VN).

**Lesions**

Su [47] Yes 78/110 96 87 93 92 92 Tischendorf [52] Yes 99/200 94 89 94 89 92 East [51] a Yes 30/33 77-91 50-60 - - 69-81

**Author Mag Patients/**

NBI has also been evaluated to diagnose early colorectal neoplasia and submucosal inva‐ sion. Katagiri et al. [61] used the capillary pattern classification in colon adenomas. Those showing CP type III harbored HGD or invasive cancer. In a recent report this group further developed this classification expanding CP type III in group IIIA (visible microvascular ar‐ chitecture and high microvessel density with lack of uniformity, branching and curtailed ir‐ regularity) and group IIIB (nearly avascular or loose microvascular area). This detailed classification allowed differentiation between lesions with Sm1 submucosal invasion from Sm2-Sm3 with a sensitivity, specificity and diagnostic accuracy of 84.8%, 88.7% and 87.7% respectively [62]. Hirata et al.[50] found that the accuracy of diagnosis of submucosal mas‐ sive invasion on the basis of thick and severely irregular vascular pattern was 100%. Kanao et al. [55] used a combination of capillary pattern and pit pattern and showed that lesions with irregular microvessels with variable sizes and distribution, and pit absence with avas‐ cular areas harbored more often massive submucosal invasion.

**Author Predictive of adenoma Predictive of hyperplastic**

Fine capillary network alone but absent mucosal

http://dx.doi.org/10.5772/53205

185

Circular pattern with dots (central dark area

Pattern of black dots surrounded by white

Thin blood vessels coursing across polyp surface, and

Non-neoplastic pit pattern (circular pit pattern)

surrounded by a lighter area)

Bland, featureless appearance

not surrounding pits

No vascular markins

pattern

In vivo Optical Diagnosis of Polyp Histology: Can We Omit Pathological Examination of Diminutive Polyps?

Round/oval pattern (dark outer and a lighter

central area) Tubulogyrus pattern

Short thick blood vessels

Central brown depression

rectangles, pentagons, etc.

Sikka [69] Neoplastic pit pattern (elongation of crypts, cerebriform pattern) Increased vascular markins

**Table 4.** Prediction of histology using NBI without magnification

*2.2.6. Prediction of histology of diminutive polyps*

Tubular or oval pits, variable size pits

Straight blood vessels around pits forming

Rogart [68] Modified Kudo´s classification Vascular color intensity (light, medium, dark)

The group of the Indiana University has very recently designed a simple classification for determination of polyp histology (NICE classification) and has validated it for its use by ex‐ perienced and non-experienced examinators (table 5) [70]. Further studies are needed to

Some authors have evaluated de diagnostic accuracy of NBI on diminutive polyps showing similar results to those on polyps of any size. In a study by Rex [67] the sensitivity of NBI in diagnosing adenomas was 92%, specificity 87%, PPV was 88%, NPV 91% and accuracy 89%. Grading the confidence on the endoscopic diagnosis in high and low, high confidence pre‐ dictions of adenomas were correct in 92% of polyps and in 91% of ≤ 5 mm polyps. The equivalent figures for hyperplastic prediction were 95%. The same group evaluated the per‐ formance of NBI in real time for distal colorectal polyps, and showed a sensitivity of 96%, a specificity of 99.4%, and NPV and PPV of 99.4% and 96% respectively [71]. The authors con‐ cluded that NBI is sufficiently accurate to allow distal hyperplastic polyps to be left in place without resection and small, distal adenomas to be discarded without pathologic assess‐ ment. In the study of Henry et al. [54] the sensitivity for predicting histology was 87%, spe‐ cificity was 93%, PPV was 89%, NPV was 91% and overall accuracy was 90%. Paggy et al. [72] found similar results both in the whole group of < 10 mm polyps and in diminutive pol‐ yps. Other authors have not showed as good results [56,73]. The most recent report using the NICE classification found an accuracy of 89%, sensitivity of 98% and a NPV of 95%. In conclusion, diagnostic accuracy of endoscopic prediction of histology of diminutive polyps

evaluate the reproducibility of this classification in real-time endoscopy.

seems equivalent to that of larger polyps, at least in expert hands.

Rex [67] Overall brown color

Rastogi [49, 65,

66]

#### *2.2.4. NBI compared with other diagnostic modalities*

NBI has been compared with other image enhancing technologies, most frequently with chromoendoscopy. Overall, the diagnostic accuracy of NBI is better than that of convention‐ al colonoscopy and equivalent to that of chromoendoscopy (figure 3) [46,47,52], especially if vascular assessment rather than pit pattern is used [51].

**Figure 3.** Invasive carcinoma in a depressed lesion observed with white light (a), NBI (b), and chromoendoscopy (c)

Four recent studies perform an evaluation of endoscopic trimodal imaging (high-resolution endoscopy, autofluorescence imaging and NBI) for colonic polyp characterization. Three studies from the same group show a poor diagnostic accuracy for NBI without magnifica‐ tion and autofluorescence with similar sensitivity but worse specificity [59,63,64]. Ignjatovic et al. [56] reported that NBI with magnification appeared to have the best accuracy, albeit modest and not adequate for in vivo diagnosis.

#### *2.2.5. NBI without optical magnification*

Most of the studies on prediction of histology using NBI have been carried out in Japan us‐ ing Olympus equipments with optical magnification (Lucera), a feature not included in high-resolution systems (Exera) available in the USA and in continental Europe. Most of the capillary pattern descriptions or classifications have been designed using optical magnifica‐ tion, therefore are not directly applicable to high-resolution examinations. That is also the case for the Kudo´s pit pattern classification.

The results of NBI without optical magnification in predicting histology are variable with authors showing an accuracy similar to that of optical magnification NBI and authors ob‐ taining worse results [56]. Again, different definitions for a vascular pattern typical of ade‐ noma (table 4) may account for these discrepancies. None of these classifications have been appropriately validated and its reproducibility in different clinical settings is unknown.


**Table 4.** Prediction of histology using NBI without magnification

sive invasion on the basis of thick and severely irregular vascular pattern was 100%. Kanao et al. [55] used a combination of capillary pattern and pit pattern and showed that lesions with irregular microvessels with variable sizes and distribution, and pit absence with avas‐

NBI has been compared with other image enhancing technologies, most frequently with chromoendoscopy. Overall, the diagnostic accuracy of NBI is better than that of convention‐ al colonoscopy and equivalent to that of chromoendoscopy (figure 3) [46,47,52], especially if

**Figure 3.** Invasive carcinoma in a depressed lesion observed with white light (a), NBI (b), and chromoendoscopy (c)

Four recent studies perform an evaluation of endoscopic trimodal imaging (high-resolution endoscopy, autofluorescence imaging and NBI) for colonic polyp characterization. Three studies from the same group show a poor diagnostic accuracy for NBI without magnifica‐ tion and autofluorescence with similar sensitivity but worse specificity [59,63,64]. Ignjatovic et al. [56] reported that NBI with magnification appeared to have the best accuracy, albeit

Most of the studies on prediction of histology using NBI have been carried out in Japan us‐ ing Olympus equipments with optical magnification (Lucera), a feature not included in high-resolution systems (Exera) available in the USA and in continental Europe. Most of the capillary pattern descriptions or classifications have been designed using optical magnifica‐ tion, therefore are not directly applicable to high-resolution examinations. That is also the

The results of NBI without optical magnification in predicting histology are variable with authors showing an accuracy similar to that of optical magnification NBI and authors ob‐ taining worse results [56]. Again, different definitions for a vascular pattern typical of ade‐ noma (table 4) may account for these discrepancies. None of these classifications have been appropriately validated and its reproducibility in different clinical settings is unknown.

cular areas harbored more often massive submucosal invasion.

*2.2.4. NBI compared with other diagnostic modalities*

184 Colonoscopy and Colorectal Cancer Screening - Future Directions

modest and not adequate for in vivo diagnosis.

case for the Kudo´s pit pattern classification.

*2.2.5. NBI without optical magnification*

vascular assessment rather than pit pattern is used [51].

The group of the Indiana University has very recently designed a simple classification for determination of polyp histology (NICE classification) and has validated it for its use by ex‐ perienced and non-experienced examinators (table 5) [70]. Further studies are needed to evaluate the reproducibility of this classification in real-time endoscopy.

#### *2.2.6. Prediction of histology of diminutive polyps*

Some authors have evaluated de diagnostic accuracy of NBI on diminutive polyps showing similar results to those on polyps of any size. In a study by Rex [67] the sensitivity of NBI in diagnosing adenomas was 92%, specificity 87%, PPV was 88%, NPV 91% and accuracy 89%. Grading the confidence on the endoscopic diagnosis in high and low, high confidence pre‐ dictions of adenomas were correct in 92% of polyps and in 91% of ≤ 5 mm polyps. The equivalent figures for hyperplastic prediction were 95%. The same group evaluated the per‐ formance of NBI in real time for distal colorectal polyps, and showed a sensitivity of 96%, a specificity of 99.4%, and NPV and PPV of 99.4% and 96% respectively [71]. The authors con‐ cluded that NBI is sufficiently accurate to allow distal hyperplastic polyps to be left in place without resection and small, distal adenomas to be discarded without pathologic assess‐ ment. In the study of Henry et al. [54] the sensitivity for predicting histology was 87%, spe‐ cificity was 93%, PPV was 89%, NPV was 91% and overall accuracy was 90%. Paggy et al. [72] found similar results both in the whole group of < 10 mm polyps and in diminutive pol‐ yps. Other authors have not showed as good results [56,73]. The most recent report using the NICE classification found an accuracy of 89%, sensitivity of 98% and a NPV of 95%. In conclusion, diagnostic accuracy of endoscopic prediction of histology of diminutive polyps seems equivalent to that of larger polyps, at least in expert hands.


**3. Conclusion**

be supported by Professional Societies.

Address all correspondence to: mbustamantebalen@gmail.com

EndoscopyUnit. University Hospital La Fe.Valencia, Spain

to take this step forward.

Marco Bustamante-Balén\*

**Author details**

**References**

10-29.

138, 2177-90.

papers/DE/de050402.pdf.

New image-enhancing technologies may allow in vivo histological assessment of colorectal polyps, avoiding the need to pathological evaluation of all resected polyps. This would rep‐ resent substantial savings and a more direct planning of surveillance intervals [81]. Howev‐ er, there are several steps to achieve before the resect and discard strategy is widely implemented. First a more simple, reproducible and validated way of characterize colon le‐ sions is needed, especially in community practice. Learning the technique is also crucial be‐ cause when learning curve is achieved NBI performs significantly better [68]. Moreover, implementing PIVI guidelines [33] implies accepting a 10% rate of false negative when in vivo assessing histology of rectal polyps. Endoscopists may feel more comfortable with a much lower rate before leaving polyps behind. Finally, if in vivo histology is applied in dai‐ ly practice this represents a turning point in the management of colon polyps, which must

In vivo Optical Diagnosis of Polyp Histology: Can We Omit Pathological Examination of Diminutive Polyps?

http://dx.doi.org/10.5772/53205

187

In vivo histology seems here to stay, but we are still at the beginning of the way. Improve‐ ment in equipments and development of new technologies will help the medical community

[1] Siegel, R., Naishadham, D., & Jemal, A. (2012). Cancer Statistics . *CA Cancer J Clin*, 62,

[2] Ferlay, J., Autier, P., Boniol, M., Heanue, M., Colombet, M., & Boyle, P. (2007). Esti‐ mates of the cancer incidence and mortality in Europe in 2006. *Ann Oncol*, 18, 581-92.

[3] Gellad, Z. F., & Provenzale, D. (2010). Colorectal cancer: national and international perspective on the burden of the disease and public health impact. *Gastroenterology*,

[4] Oliva, J., Lobo, F., López-Bastida, J., Zozaya, N., & Romay, R. Pérdidas de productivi‐ dad laboral ocasionadas por los tumores en España. http://docubib.uc3m.es/working‐

**Table 5.** The NBI International colorectal endoscopic (NICE) classification

#### *2.2.7. Learning NBI. Does expertise matters?*

Most of the published studies have been performed by experts endoscopists, both in Japan and in Western countries. Reliable information about reproducibility of this results is lack‐ ing. Moreover, the overall accuracy in prediction of histology es markedly influenced by expertise in NBI interpretation, as has been shown in a study performed in a non academic setting in which sensitivity for high-confidence prediction was 77% and specificity 78% [73]. Experts have been shown to perform better than non-experts and with a higher interobserv‐ er agreement [74]. Fortunately, NBI interpretation of histology can be easily learned. Several studies have shown significant improvements in diagnostic accuracy and in interobserver agreement after following a computer-based training module [75] or a short teaching ses‐ sion [76].

#### **2.3. Fujinon intelligent color enhancement system (FICE) and i-Scan**

FICE also narrowes the bandwidth of light components using a computed spectral estima‐ tion technology that aritmetically processes the reflected photons to reconstitute virtual im‐ ages for a choice of different wavelenghts [77]. Therefore, it no depends on optical filters to modify the image. There are less studies using FICE or i-Scan than NBI but its accuracy seems broadly similar.

In the study by Pohl et al. [77] FICE (with set 4 activated) was used to identify the pit pattern and the vascular pattern intensity in a similar way to NBI. The sensitivity and specificity of FICE for the prediction of adenoma was 93.2% and 61.2%, figures similar to those of chro‐ moendoscopy. Parra et al. [78] showed that FICE performance in predicting histology was inferior to that of chromoendoscopy with magnification. Kim et al. [80] reported that FICE with magnification was better than without magnification especially for diminutive polyps [79]. Regarding i-Scan, a study compared this technology with NBI for histology prediction of diminutive polyps and showed a similar performance with good agreement between the two modalities (kappa index > 0.7).
