Electronic nose to distinguish bladder cancer by urinary odour feature: A pilot study

Abstract

BACKGROUND:

No study has yet investigated the use of electronic nose (eNose) technology to reveal pattern recognition of urological diseases, including bladder cancer.

OBJECTIVE:

We sought to determine the diagnostic performance of the eNose in recognizing urinary odour in patients with bladder cancer.

METHODS:

The eNose is a commercially available model equipped with two sensors. The angle of the two sensors ( $\theta$ ) depends on the kinds of chemical substances, thus defining $\theta$ as the feature of odour. Quantity of odour is the number of $\theta$ detected during a measurement. Urine samples were from 36 untreated patients with bladder cancer, 29 with urolithiasis, 10 with urinary tract infection (UTI), and 27 healthy volunteers.

RESULTS:

Based on ROC analysis of the quantity in patients with bladder cancer, an optimal cut-off value for $\theta$ of 49, 48, and 55 was applied to compare with samples from the healthy volunteer, urolithiasis and UTI groups, respectively. There were significantly differences between bladder cancer and the other conditions using these specific points ( $p<$ 0.0001, respectively). The resulting diagnostic sensitivity was 61.4%, 45.6%, and 60.8%, and specificity was 52.8%, 68.4%, and 90.2%, respectively. The AUC for bladder cancer was 0.565, 0.548, and 0.909, respectively.

CONCLUSION:

The eNose is a small, portable, rapid, low cost, and noninvasive instrument for distinguishing bladder cancer from other benign conditions.

Keywords

Bladder cancer electronic nose urine odour

1. Introduction

Urine has many components and varies depending on multiple factors, including age, hormonal condition, physical activity, food consumption and condition of the urinary tract system [1, 2, 3, 4, 5]. The identification of biomarkers in urine for early and non-invasive diagnosis purposes has been the subject of much research [6, 7, 8], and these studies revealed the chemical complexity of urine components applied to liquid urine [9] and even to its gaseous headspace [3]. The analysis of urine odour for cancer diagnosis has recently become a subject of great research interest [10, 11].

Since the first case study reported the olfactory detection of cancer [12], experimental studies using trained dogs in the detection of cancer have shown promising preliminary outcomes [13, 14]. Trained dogs were able to detect prostate cancer using urine samples with high sensitivity and specificity, leading to considerable attention [10]. However, there were possible issues regarding costs, space, duration and reproducibility, which was affected by the canine olfactory skills depending on the performance of sniffer dogs.

The electronic nose (eNose) is an instrument that emulates the human olfactory system [15]. With development of key technology such as odour handling and delivery system, odour sensor, signal processing, pattern analysis, eNose has been used in numerous applications, including in the food and beverage quality assessment and in the detection of explosive and chemical agents [16, 17, 18, 19]. The interaction of volatile organic compounds with an array of partial selective chemical sensors that may react like the olfactory receptors in the human nose results in a conductance change in the sensors, which is transmitted to a processor. Studies have shown that the use of eNose technology has many useful potential applications for medical field, including the diagnosis of bile acid diarrhoea, tuberculosis, renal dysfunction, urinary tract infection, and cancers [20, 21, 22, 23, 24]. In addition, great achievement to show the usefulness of eNose for the detection of lung cancer was reported [25].

Although several studies have reported comparisons on the use odour for the detection of bladder cancer using urine samples from cancer patients and health controls [14, 21, 26, 27, 28], to our knowledge, no study has yet investigated the use of eNose technology to reveal pattern recognition of urological diseases, including bladder cancer. In this retrospective pilot study, we sought to determine the diagnostic performance of the eNose in recognizing urinary odour in patients with bladder cancer.

2. Materials and methods

2.1 Study population

Urine samples selected from 36 untreated patients with bladder cancer of various stages and grades,29 patients with urolithiasis, 10 patients with urinary tract infection (UTI), and 27 healthy volunteers were obtained at Kitasato University Hospital. There were no samples from patients with chronic renal failure, diabetes mellitus or other malignancies. All bladder cancer patients were treated with transurethral resection or radical cystectomy. Histological cancer types other than urothelial carcinoma were also excluded. The 2002 TNM classification system developed by the American Joint Committee on Cancer and the Union for International Cancer Control (UICC) [29], was used for pathological staging (Ta: 4 patients; Tis: 2; T1: 7; T2: 9; T3: 10; T4: 4), and the 1973 World Health Organization classification was used for pathological grading (grade 1: 4 patients; grade 2: 12; grade 3: 20). The bladder cancer patients consisted of 27 male and 9 female subjects, with a median age of 74 years (range, 52–85; mean, 73.0). None of the patients had received chemotherapy or radiation before surgery. Patients with urolithiasis including renal and ureteral stones were diagnosed by imaging. Patients suffered from UTI consisted of acute pyelonephritis and cystitis. Healthy volunteers did not have urogenital cancers, other malignancies, uropathological conditions or other diseases.

Urine samples were collected at the time of spontaneous urination before treatment. Immediately after sample collection, urine samples were centrifuged at 1,000 $\times$ $g$ for 10 minutes at 4 ${}^{\circ}$ C to remove cellular components and stored at $-$ 80 ${}^{\circ}$ C until use.

This study was approved by the Ethics Committee of Kitasato University School of Medicine (approval number: C09-504, B17-164). All patients were informed of the aim of the study and gave consent for the use of their samples.

2.2 Olfactory measurement

Figure 1.

Urine odour feature. The output of A sensor (SA ${}_{\text{out}}$ : large molecules) determines position in the $X$ -axis and that of the B sensor (SB ${}_{\text{out}}$ : small molecules) in the $Y$ -axis, resulting in the odour vector (V). The strength of the odour is the length of V, and the angle between the results of the two sensors ( $\theta$ ) defines the odour feature.

Figure 2.

The strength of volatile NH ${}_{3}$ and H ${}_{2}$ S from urine. Normal: healthy volunteers; BT: bladder cancer; stone: urolithiasis; UTI: urinary tract infection. Standard error of the mean is indicated by the bars.

The eNose used in this study is a commercially available model (e-nose ${}^{\text{\textregistered}}$ integral, e-nose Instruments Co., Ltd. Japan) equipped with four electrochemical sensors with metal oxide semiconductor. Metrological parameters of this sensor are sensitivity. The A sensor is mainly sensitive to volatile large molecules, like aromatic hydrocarbons, and the B sensor is mainly sensitive to volatile small molecules, like alcohol. The output of A sensor (SA ${}_{\text{out}}$ ) determines position in the $X$ -axis and that of B sensor (SB ${}_{\text{out}}$ ) in the $Y$ -axis (Fig. 1), resulting in an odour vector (V). The strength of the odour is the length of the V determined by the two sensors. The angle of the two sensors ( $\theta$ ), that is the slope of the odour, depends on the kinds of chemical substances and their composition, so $\theta$ can be defined as the feature of odour. The sensors do not specify molecules but produce a characteristic smell print of the sample. The quantity of odour is defined by the number of the features of odour present during a measurement. That is sensor counted the number of each odour factor during a measurement. The strength of volatile NH ${}_{3}$ (C sensor) and H ${}_{2}$ S (D sensor) were also measured. Ambient air was used as carrier gas. The device was connected to a Windows-based portable computer with software for monitoring and logging the data.

The measurement equipment consisted of a polysty-rene cell culture flask (25 ml) with air filter cap covered by paraffin film. An 18 G intravenous cannula was connected to a silicon tube, and the afferent side was connected to the eNose. These equipment were used only once. Before analysis, samples were slowly thawed and gently mixed. Urine (5 ml) was put into a culture flask. The sample in the flask was heated to 45 ${}^{\circ}$ C by a water bath and maintained for 3 minutes. While the culture flask was standing obliquely in the water bath, the paraffin film was removed, and the 18 G intravenous cannula was inserted into air filter cap without touching the urine sample. After the sample air entered into the silicon tube for up to 1 minute, each measurement lasted approximately 1 minute and reached a plateau, depending on the samples. For the next sample, a recovery period of at least 10 minutes was made using activated carbon to equalize the eNose and avoid carryover. We performed several pilot experiments [30, 31], and these incubation times and intervals were determined to be sufficient for measuring urinary odour (data not shown). Every sample was run in duplicate and the mean calculated for data analysis.

2.3 Statistical analysis

For this analysis, gender, age ( $\leqq$ 65 vs $>$ 65, 7 and 29 patients with bladder cancer, respectively), tumour stage ( $\leqq$ T1 versus $\geqq$ T2) and grade (grade 1 and 2 vs grade 3) were evaluated as dichotomized variables. Significant differences between the odour outcomes and clinicopathologic factors were assessed by the Mann-Whitney $U$ test and the Kruskal-Wallis test. The area under the curve (AUC) and best cut-off point were calculated employing receiver-operating characteristic curve (ROC) analysis. A $p$ -value of less than 0.05 was used to determine the level of significance. All statistical analyses were performed using StatFlex software version 6.0 (Artech Co., Ltd., Osaka, Japan).

3. Results

3.1 Strength of volatile NH ${}_{3}$ and H ${}_{2}$ S

The strength of volatile NH ${}_{3}$ and H ${}_{2}$ S from urine are shown in Fig. 2. Volatile NH ${}_{3}$ was highest in patients with UTI ( $p=$ 0.016). There was no significant difference among the groups in terms of volatile H ${}_{2}$ S. In patients with bladder cancer, age, gender, pathological stage and grade were not linked with any significant differences in volatile NH ${}_{3}$ and H ${}_{2}$ S from urine samples.

Table 1
AUC and best cut-off points for urine odour quantity and feature based on ROC analysis.

	Cut off ( $\theta$ )	Sensitivity (%)	Specificity (%)	AUC	95% CI	$P$ value*
BT-normal	49	61.4	52.8	0.565	0.419–0.711	$<$ 0.0001
BT-stone	48	45.6	68.4	0.548	0.459–0.636	$<$ 0.0001
BT-UTI	55	60.8	90.2	0.909	0.897–0.920	$<$ 0.0001
normal-UTI	54	81.7	88.9	0.927	0.916–0.937	$<$ 0.0001
normal-stone	47	35.9	74.5	0.508	0.317–0.646	0.4
UTI-stone	55	60.8	87.9	0.897	0.885–0.909	$<$ 0.0001

AUC: area under curve; ROC: receiver-operating characteristic curve; CI: confidence interval; normal: healthy volunteer; BT: bladder cancer; UTI: urinary tract; * Mann-Whitney $U$ test.

Figure 3.

The association of urine odour quantity and feature. The results of odour feature ( $\theta$ ) determine position in the $X$ -axis and odour quantity in the $Y$ -axis. Normal: healthy volunteers; BT: bladder cancer; stone: urolithiasis; UTI: urinary tract infection.

3.2 Association of quantity with feature of odour

The results of the feature of odour ( $\theta$ ) determine position in the $X$ -axis and that of the quantity of odour in $Y$ -axis, resulting in a distribution of odour features (Fig. 3). These data averaged the number for each $\theta$ from all samples in each condition. The peaks of $\theta$ varied between the groups. Bladder cancer and UTI showed two different peaks in odour quantity. Urolithiasis revealed several peaks. Healthy volunteers demonstrated only one peak.

Based on ROC analysis of the odour of bladder cancer after checking the number of odour peaks, an optimal cut-off value for $\theta$ of 49, 48, and 55 was applied to compare bladder cancer to healthy volunteer, urolithiasis, and UTI, respectively (Fig. 4). There were significantly differences between bladder cancer and the other conditions using these specific points ( $p<$ 0.0001, respectively). The resulting diagnostic sensitivity was 61.4%, 45.6%, and 60.8%, and the specificity was 52.8%, 68.4%, and 90.2%, respectively. The AUC for the odour of bladder cancer was 0.565, 0.548, and 0.909, respectively. The other ROC analyses of odour between healthy volunteers, patients with urolithiasis and UTI are shown in Table 1.

Figure 4.

Receiver-operating characteristic curve analysis according to the association of odour feature and quantity. a: bladder cancer and healthy volunteers; b: bladder cancer and urolithiasis; c: bladder cancer and urinary tract infection.

3.3 Association of strength with feature of odour

A relationship between V and $\theta$ was also examined, as well as with odour quantity (Fig. 5). There were no significant differences between patients with bladder cancer, those with urolithiasis, and healthy volunteer based on ROC analysis. Patients with UTI showed significantly strong odour compared to those with bladder cancer, those with urolithiasis, and healthy volunteers using the same analysis ( $p<$ 0.0002, respectively).

Figure 5.

The association of the urine odour strength and feature. The results of odour feature ( $\theta$ ) determines position in the $X$ -axis and odour strength (V) in the $Y$ -axis. Normal: healthy volunteers; BT: bladder cancer; stone: urolithiasis; UTI: urinary tract infection.

4. Discussion

Based on pattern recognition, the detection of various conditions from urine samples using eNose equipment is a high-throughput method. We found that the association between urine odour feature ( $\theta$ ) and quantity – but not strength (V) – was significantly different between bladder cancer and the other conditions, including urolithiasis, UTI, and healthy volunteers. This study also shows that the eNose possibly distinguished bladder cancer from the other conditions by three specific points ( $\theta$ : 48, 49 and 55) based on ROC analyses. The eNose has potential for use as a diagnostic tool because it is noninvasive, rapid, cheap, and has good reproducibility, providing a new modality in the detection of bladder cancer. Early diagnosis could lead to better appropriate and early treatment, less loss of function, and a higher survival rate.

Regarding urological oncology, bladder cancer is one of the diseases that received olfactory analysis in animal models as a diagnostic modality. One study demonstrated that dogs can be well trained to recognize bladder cancer in urine samples, showing a diagnostic success rate of 41%, compared to a 14% success rate reported for chance [14]. The other study using sniffer mice indicated that trained mice discriminated between the urinary odour of pre- and post-transurethral resection in individual patients with bladder cancer, achieving a success rate of 100% [26].

Several studies have also investigated urinary odour for the detection of bladder cancer using the eNose instead of an animal model. Bernabei et al. evaluated odour using eNose and was able to discriminate the urine samples of healthy patients from those of patients with bladder cancer, with a diagnostic accuracy of 100% [21]. Heers et al. showed that the eNose correctly detected 93.3% already confirmed transitional cell carcinoma and 86.7% of healthy controls by urine samples [27]. Another experiment reported by Weber et al. studied the interaction between volatile organic compounds in the urine sample and bladder cancer [28]. An accuracy of 70% was reached, but this was lower when regarding patients with other infection diseases. They concluded that more sophisticated pattern recognition is needed to improve the results. The response provided by the eNose may depend on a certain property of the sensor that was transduced into an electrical signal. However, based on previous reports, diagnostic accuracy in urinary odour was merely reported between bladder cancer patients and healthy volunteers. The presented results demonstrate clear pattern recognition and discrimination between patients with bladder cancer and other conditions using the feature of odour ( $\theta$ ). We provide the potential use of the eNose as a quick diagnostic instrument for the detection of urological disease related to the analysis of urine odour.

UTIs are a significant cause of morbidity, with a continuously rising number of patients. The analysis of the volatile organic compounds connected to these diseases by means of the eNose was investigated as a possible instrument of diagnosis. One study showed the detection of microbial contaminants such as Escherichia coli in urine samples using the eNose [20]. In the present study, the odour of UTI was the strongest in the groups in terms of volatile NH ${}_{3}$ and $\theta$ , which the physicians clinically recognized. This was a reasonable result because UTI is mainly caused by bacteria like E. coli or enteric and fungal pathogens. The eNose system based on $\theta$ may provide a potential avenue for the discrimination of bacterial species.

Some limitations of our study should be mentioned. First, the number of cases was retrospective and small to permit definitive conclusions on the diagnostic accuracy of the eNose, although the preliminary data appears at least promising. Second, we used urine samples stored at $-$ 80 ${}^{\circ}$ C, which may be clinically unusual condition, although one study has reported that there were no differences in terms of detection rate of bladder cancer between storage of urine samples at $-$ 20 ${}^{\circ}$ C and $-$ 80 ${}^{\circ}$ C [27]. In a next step, we need to check the dynamics of odour in fresh urinary samples prospectively. Third, we did not check combined diseases such as patients with bladder cancer and UTI or with gross haematuria and urolithiasis, and such situations are often seen in real world clinical scenarios. Based on this functioning principle of the eNose, the sensor responses produced a kind of odour fingerprint as a whole, which meant that the eNose was not able to identify single compounds. However, the eNose system using $\theta$ in the present study may be closer to the elements of urinary odour than previously published reports. Finally, urinary conditions are affected by a variety of clinical circumstances, including renal function, medicine, food consumption, and other systematic diseases. Although we selected patients with only one disease for this pilot study, these issues need to be overcome to use the eNose as a conventional tool in daily clinical practice. While the present study includes these limitations, it also clearly indicates that the eNose may be able to distinguish bladder cancer from a variety of urinary conditions by checking the association between the urine odour feature ( $\theta$ ) and quantity. Further studies are warranted as the next step for the external validation of the eNose in diagnostic assessment of bladder cancer, which is necessary to make this new technology acceptable as an efficacious system.

5. Conclusions

The eNose is a small, portable, rapid, low cost, and noninvasive instrument for distinguishing bladder cancer from other benign conditions. An analysis of the spectrum of urinary odour is possible, reinforcing its potential as a clinical tool, though results need to be validated due to the limited number of subjects involved in the study. More studies are needed to establish protocols for the use of the eNose in a clinical setting. The eNose could either become a diagnostic tool for excluding patients with urolithiasis and UTI or for selecting patients for more invasive procedures like cystoscopy.

Footnotes

Acknowledgments

This study was supported in part by JSPS KAKENHI Grant Number JP18K09206.

References

Kwak

Grigsby

C.C.

Smith

B.R.

Rizki

M.M.

and Preti

, Changes in volatile compounds of human urine as it ages: Their interaction with water, J Chromatogr B Analyt Technol Biomed Life Sci 941 (2013), 50–53.

Matsumoto

Shariat

S.F.

Casella

Wheeler

T.M.

Slawin

K.M.

and Lerner

S.P.

, Preoperative plasma soluble E-cadherin predicts metastases to lymph nodes and prognosis in patients undergoing radical cystectomy, J Urol 170 (2003), 2248–2252.

Mills

G.A.

and Walker

, Headspace solid-phase microextraction profiling of volatile compounds in urine: Application to metabolic investigations, J Chromatogr B Biomed Sci Appl 753 (2001), 259–268.

Samudrala

Geurts

Brown

P.A.

Szymanska

Mandon

Jansen

Buydens

Harren

F.J.

and Cristescu

S.M.

, Changes in urine headspace composition as an effect of strenuous walking, Metabolomics 11 (2015), 1656–1666.

Smith

Burden

Persad

Whittington

de Lacy Costello

Ratcliffe

N.M.

and Probert

C.S.

, A comparative study of the analysis of human urine headspace using gas chromatography-mass spectrometry, J Breath Res 2 (2008), 037022.

Lotan

O’Sullivan

Raman

J.D.

Shariat

S.F.

Kavalieris

Frampton

Guilford

Luxmanan

Suttie

Crist

Scherr

Asroff

Goldfischer

Thill

and Darling

, Clinical comparison of noninvasive urine tests for ruling out recurrent urothelial carcinoma, Urol Oncol 35(531) (2017), e15–e22.

Shariat

S.F.

Matsumoto

Casella

Jian

and Lerner

S.P.

, Urinary levels of soluble e-cadherin in the detection of transitional cell carcinoma of the urinary bladder, Eur Urol 48 (2005), 69–76.

Soria

D’Andrea

Pohar

Shariat

S.F.

and Lotan

, Diagnostic, prognostic and surveillance urinary markers in nonmuscle invasive bladder cancer: any role in clinical practice? Curr Opin Urol 28 (2018), 577–583.

Benkali

Marquet

Rerolle

Le Meur

and Gastinel

, A new strategy for faster urinary biomarkers identification by Nano-LC-MALDI-TOF/TOF mass spectrometry, BMC Genomics 9 (2008), 541.

10.

Cornu

J.N.

Cancel-Tassin

Ondet

Girardet

and Cussenot

, Olfactory detection of prostate cancer by dogs sniffing urine: A step forward in early diagnosis, Eur Urol 59 (2011), 197–201.

11.

Roine

Veskimae

Tuokko

Kumpulainen

Koskimaki

Keinanen

T.A.

Hakkinen

M.R.

Vepsalainen

Paavonen

Lekkala

Lehtimaki

Tammela

T.L.

and Oksala

N.K.

, Detection of prostate cancer by an electronic nose: A proof of principle study, J Urol 192 (2014), 230–234.

12.

Williams

and Pembroke

, Sniffer dogs in the melanoma clinic? Lancet 333 (1989), 734.

13.

Sonoda

Kohnoe

Yamazato

Satoh

Morizono

Shikata

Morita

Watanabe

Morita

Kakeji

Inoue

and Maehara

, Colorectal cancer screening with odour material by canine scent detection, Gut 60 (2011), 814–819.

14.

Willis

C.M.

Church

S.M.

Guest

C.M.

Cook

W.A.

McCarthy

Bransbury

A.J.

Church

M.R.

and Church

J.C.

, Olfactory detection of human bladder cancer by dogs: Proof of principle study, BMJ 329 (2004), 712.

15.

Pearce

T.C.

, Computational parallels between the biological olfactory pathway and its analogue ‘the electronic nose’: Part II. Sensor-based machine olfaction, Biosystems 41 (1997), 69–90.

16.

Kramer

K.E.

Rose-Pehrsson

S.L.

Hammond

M.H.

Tillett

and Streckert

H.H.

, Detection and classification of gaseous sulfur compounds by solid electrolyte cyclic voltammetry of cermet sensor array, Anal Chim Acta 584 (2007), 78–88.

17.

Fernandes

D.L.

and Gomes

M.T.

, Development of an electronic nose to identify and quantify volatile hazardous compounds, Talanta 77 (2008), 77–83.

18.

Wilson

and Baietto

, Applications and advances in electronic-nose technologies, Sensors 9 (2009), 5099–5148.

19.

Majchrzak

Wojnowski

Dymerski

and Namiesnik

, Electronic noses in classification and quality control of edible oils: A review, Food Chemistry 246 (2018), 192–201.

20.

Pavlou

A.K.

Magan

McNulty

Jones

Sharp

Brown

and Turner

A.P.

, Use of an electronic nose system for diagnoses of urinary tract infections, Biosens Bioelectron 17 (2002), 893–899.

21.

Bernabei

Pennazza

Santonico

Roscioni

Paolesse

Di Natale

and D’Amico

, A preliminary study on the possibility to diagnose urinary tract cancers by an electronic nose, Sens Actuator B Chem 131 (2007), 1–4.

22.

Kolk

Hoelscher

Maboko

Jung

Kuijper

Cauchi

Bessant

van Beers

Dutta

Gibson

and Reither

, Electronic-nose technology using sputum samples in diagnosis of patients with tuberculosis, J Clin Microbiol 48 (2010), 4235–4238.

23.

Covington

J.A.

Westenbrink

E.W.

Ouaret

Harbord

Bailey

O’Connell

Cullis

Williams

Nwokolo

C.U.

Bardhan

K.D.

and Arasaradnam

R.P.

, Application of a novel tool for diagnosing bile acid diarrhoea, Sensors 13 (2013), 11899–11912.

24.

Bax

Taverna

Eusebio

Sironi

Grizzi

Guazzoni

and Capelli

, Innovative diagnostic methods for early prostate cancer detection through urine analysis: A review, Cancers 10 (2018), 123.

25.

D’Amico

Pennazza

Santonico

Martinelli

Roscioni

Galluccio

Paolesse

and Di Natale

, An investigation on electronic nose diagnosis of lung cancer, Lung Cancer 68 (2010), 170–176.

26.

Sato

Katsuoka

Yoneda

Nonomura

Uchimoto

Kobayakawa

and Mizutani

, Sniffer mice discriminate urine odours of patients with bladder cancer: A proof-of-principle study for non-invasive diagnosis of cancer-induced odours, Sci Rep 7 (2017), 14628.

27.

Heers

Gut

J.M.

Hegele

Hofmann

Boeselt

Hattesohl

and Koczulla

A.R.

, Non-invasive detection of bladder tumors through volatile organic compounds: A pilot study with an electronic nose, Anticancer Res 38 (2018), 833–837.

28.

Weber

C.M.

Cauchi

Patel

Bessant

Turner

Britton

L.E.

and Willis

C.M.

, Evaluation of a gas sensor array and pattern recognition for the identification of bladder cancer from urine headspace, Analyst 136 (2011), 359–364.

29.

Greene

F.L.

Page

D.L.

Fleming

I.D.

et al., AJCC Cancer Staging Manual, 6th ed, New York, NY: Springer 2002.

30.

Abbott

S.M.

Elder

J.B.

Spanel

and Smith

, Quantification of acetonitrile in exhaled breath and urinary headspace using selected ion flow tube mass spectrometry, Int J Mass Spectrom 228 (2003), 655–665.

31.

Majchrzak

Wojnowski

Piotrowicz

Gebicki

and Namiesnik

, Sample preparation and recent trends in volatolomics for diagnosing gastrointestinal diseases, Trends Anal Chem 108 (2018), 38–49.

Electronic nose to distinguish bladder cancer by urinary odour feature: A pilot study

Abstract

BACKGROUND:

OBJECTIVE:

METHODS:

RESULTS:

CONCLUSION:

Keywords

1. Introduction

2. Materials and methods

2.1 Study population

2.2 Olfactory measurement

3. Results

3.1 Strength of volatile NH 3 and H 2 S

Table 1 AUC and best cut-off points for urine odour quantity and feature based on ROC analysis.

5. Conclusions

Footnotes

Acknowledgments

References

3.1 Strength of volatile NH ${}_{3}$ and H ${}_{2}$ S

Table 1
AUC and best cut-off points for urine odour quantity and feature based on ROC analysis.