Automatic classification of lymphoma lesions in FDG-PET-Differentiation between tumor and non-tumor uptake.

INTRODUCTION: The automatic classification of lymphoma lesions in PET is a main topic of ongoing research. An automatic algorithm would enable the swift evaluation of PET parameters, like texture and heterogeneity markers, concerning their prognostic value for patients outcome in large datasets. Moreover, the determination of the metabolic tumor volume would be facilitated. The aim of our study was the development and evaluation of an automatic algorithm for segmentation and classification of lymphoma lesions in PET.
METHODS: Pre-treatment PET scans from 60 Hodgkin lymphoma patients from the EuroNet-PHL-C1 trial were evaluated. A watershed algorithm was used for segmentation. For standardization of the scan length, an automatic cropping algorithm was developed. All segmented volumes were manually classified into one of 14 categories. The random forest method and a nested cross-validation was used for automatic classification and evaluation.
RESULTS: Overall, 853 volumes were segmented and classified. 203/246 tumor lesions and 554/607 non-tumor volumes were classified correctly by the automatic algorithm, corresponding to a sensitivity, a specificity, a positive and a negative predictive value of 83%, 91%, 79% and 93%. In 44/60 (73%) patients, all tumor lesions were correctly classified. In ten out of the 16 patients with misclassified tumor lesions, only one false-negative tumor lesion occurred. The automatic classification of focal gastrointestinal uptake, brown fat tissue and composed volumes consisting of more than one tissue was challenging.
CONCLUSION: Our algorithm, trained on a small number of patients and on PET information only, showed a good performance and is suitable for automatic lymphoma classification.

Introduction

Fluorodeoxyglucose positron emission tomography (PET) is a well-established, sensitive method for the assessment of Hodgkin lymphoma (HL) and glucose-avid Non-HL [1-4]. The initial PET scan is important for exact staging and radiotherapy planning [5, 6].

Lymphoma represent a systemic disease in which multifocal involvement pattern is common [7, 8]. The precise detection of all initial lymphoma lesions is crucial for correct treatment stratification [9, 10]. The differentiation between lymphoma uptake and non-tumor uptake in PET can be challenging [11]. Reasons are the multifocal lymphoma involvement pattern with large variability in terms of size, localization and tracer uptake, as well as the complex physiological FDG distribution with variable uptake behavior, e.g. of the heart, the kidneys and the skeleton. In addition, inflammatory foci, bone marrow activation and activated brown fat tissue may hamper the PET evaluation. The assessment of PET scans of lymphoma patients is feasible for experienced nuclear physicians but quite challenging for automatic algorithms.

The initial metabolic tumor volume was reported to be an independent prognostic factor in lymphoma patients [12]. The prognostic value of further PET parameters, like texture and heterogeneity markers is subject of current research [13-17]. Large numbers of PET datasets are required to evaluate the prognostic value of different parameters. Manual detection and segmentation of lymphoma lesions is very time consuming and therefore not suitable for large datasets. Algorithms have to be developed and tested, enabling the automatic segmentation of lymphoma lesions. Moreover, an automatic classification of non-tumor lesions in different tissue categories could support dosimetry calculations.

The aim of our study was to develop and evaluate an automatic algorithm for segmentation and classification of lymphoma lesions in PET. In addition, we aimed at an optimized differentiation of the various non-tumor tissues.

Methods

Patients

Our study included pre-treatment PET scans from HL patients from the EuroNet-PHL-C1 (C1) trial (EudraCT number NCT00433459) [18]. PET imaging was performed in multiple PET centers, following the imaging recommendations of the C1 study protocol.

Ethical approval was granted by the ethics committee of the University of Leipzig (132/19-ek). C1 study patients and/or their guardians gave written informed consent and acknowledged transfer and storage of their imaging data on the Pediatric-Hodgkin-Network server [19]. For our evaluation, all PET scans were completely anonymized. Thus, the ethics committee waived the requirement for additional informed consent.

Data preparation and transfer

In preparation of an evaluation regarding prognostic PET parameters for relapse, each initial PET scan from the C1 study was anonymized and labeled with two attributes only. The first attribute represented the PET result of the patient after two cycles of chemotherapy (“PET-positive” or “PET-negative”). The second attribute indicated whether a relapse occurred later on or not (“relapse” or “non-relapse”). Thus, each anonymized PET scan could be assigned to one of four subgroups (PET-positive relapse, PET-positive non-relapse, PET-negative relapse, PET-negative non-relapse). All anonymized PET scans were transferred via secured data connection from the Pediatric-Hodgkin-Network server to the Institute of Computer Science at the Martin-Luther-University of Halle (Saale), Germany for further evaluation.

Preprocessing

For all PET scans, the intensity values were smoothed for denoising using a Gauss filter with σ = 2.0. Thereafter, SUV values were calculated and coronal SUV projections were reconstructed for visual plausibility check. From the valid scans, 15 PET scans from each subgroup, a total of 60 scans, were drawn randomly for this study.

Segmentation

First, all voxels with an SUV below 2.5 were excluded. Second, a watershed algorithm was used for segmentation of the SUV dataset [20]. This algorithm uses the concept that individual voxels belong to certain catchment basins to define different volumes. Watersheds represent borders between these basins. To detect watersheds, the flooding of the topography is simulated. The algorithm stops when all voxels are assigned to either one catchment area or a watershed. Third, neighboring volumes were merged based on similar texture. Specifically, for each volume co-occurrence matrices were computed using a distance of one for all directions and the average of the resulting contrast, entropy and inverse difference moment was determined. Two volumes were merged if the absolute differences of their aggregated texture measures was smaller than a threshold of two.

Volumes smaller than 2.0 ml were excluded from further analysis [12, 21].

Automatic cropping algorithm

For further standardization, each PET scan was cropped along the z-axis. An heuristic algorithm was developed, cropping the PET scan at brain and urinary bladder level. An illustrated description is given in the supplementary material. Segmented volumes were not considered for further analysis if they overlapped the boundaries or laid outside.

Manual classification

All segmented volumes were manually classified by a nuclear physician with ten years of expertise in HL evaluation. Each volume was assigned to one of the following 14 categories: tumor, composed volumes of tumor and non-tumor tissue (T+NT), brain, skeleton, head-and-neck-region, heart, right kidney, left kidney, liver, gastrointestinal tract, genital organs, urinary bladder, activated brown fat tissue and composed volumes consisting of more than one non-tumor tissue (NT+NT).

Feature computation and normalization

In our study, 31 features were used for classification: 19 SUV-based features, six shape-based features and six spatial location features (Table 1). These features had to be normalized to be comparable between scans, since patients differed in their constitutions and PET scanner in their resolutions. Our approach was to compute the features in the original PET scan and subsequently transform them to a standardized size and scanner resolution.

Table 1

Features of the segmented volumes used for classification.

Features for Classification
SUV-based features		Shape-based features	Spatial location features
Maximum SUV	Mean absolute deviation of SUV	Area surface ratio	Direction of major axis (x)
Minimium SUV	Root mean squared of SUV	Surface difference2	Direction of major axis (y)
Mean SUV	Standard deviation of SUV	Scaled surface area	Direction of major axis (z)
Median SUV	Skewness of SUV	Scaled volume	Location of centorid (x)
Range of SUV	Coarseness of NGTDM	Compactness	Location of centorid (y)
Variance of SUV	Contrast of NGTDM	Maximum diameter	Location of centorid (z)
Energy of SUV	Busyness of NGTDM
Entropy of SUV	Complexity of NGTDM
Kurtosis of SUV	Outside difference1
Uniformity of SUV

1) Mean difference of SUV values between volume and surrounding.

2) Estimated average ratio of eucliden vs surface distance of surface voxel pairs.

Abbreviations: SUV—Standard uptake value, NGTDM—Neighboring gray tone difference matrix.

Automatic classification

For automatic classification of the segmented volumes, the random forest method [22] was applied. A random forest consists of several hundred up to thousands of decision trees. Each tree consists of a hierarchy of binary decisions. With every decision it was checked, if a specific feature of the volume exceeded a cut-off value or not (e.g. Is the maximum SUV above 4.0?). After navigating through the whole tree up to the leaf, the given volume was classified into the category associated with this leaf. Each volume was classified using all decision trees. The final classification result represented the majority vote of all trees.

A nested cross validation was used for training of the random forest. For ease of presentation, only an overview is given here, a detailed description can be found in the supplementary material. The whole dataset was split into three parts: a training set, a validation set and a test set. Using the training set, the random forest was trained. All decision trees of the forest were trained independently. The validation set was used to find the best combination of hyperparameters. Hyperparameters in our study were: the maximum depth of each decision tree and the number of selectable features for each decision. The trained random forest with the best hyperparameter combination was applied to the test set. This procedure was repeated until each part of the entire data set was once a test data set to derive an unbiased estimate of the performance.

Performance evaluation

The results of the automatic cropping algorithm were evaluated visually. Nested cross validation with random data partitioning and ten repetitions was used for performance evaluation of the automatic classification. The classification results were averaged across the ten repetitions and subsequently rounded to integral numbers. For the binary decision between tumor and non-tumor, T+NT volumes were considered tumor lesions since our aim was to include all tumor lesions. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and F1-score (F1 = 2 / (sensitivity-1 + PPV-1)) were calculated.

A flowchart of the methodical workflow is given in Fig 1.

Fig 1

Methodical workflow of the classification algorithm.

Results

Evaluation of the automatic cropping algorithm

The automatic cropping at brain level was successful in 50/60 patients. In nine out of the ten patients with incorrect cropping, the brain was not included in the PET acquisition. As a result, the scan was cropped at the salivary glands in six patients and at a tumor lesion in the neck in three patients. In one patient, the PET scan was cropped at a tumor lesion in the neck although the brain was visible. Due to incorrect cropping in the upper part of the scan, five tumor lesions in the neck were excluded from further evaluation.

The automatic cropping at bladder level was successful in 57/60 patients. In one patient, the scan was cropped above the bladder at the iliac crest, in one patient below the bladder at the testicles and in one patient no lower boundary was defined. However no tumor lesion got lost on the basis of incorrect cropping at the lower part of the scan. In two patients, tumor lesions were excluded from further evaluation since they were located below bladder level—inguinal lymph nodes in one patient and one skeletal lesion in the right femur in another patient.

Overall, 853 segmented volumes were equal or larger than 2.0 ml and located completely within the cropping boundaries and therefore considered for further evaluation.

225 tumor lesions and 21 composed T+NT volumes were manually classified, an average of 4.1 tumor lesions per patient. Tumor lesions occurred in lymph nodes above and below the diaphragm in 59 and 17 patients, respectively. Extra-nodal lesions occurred in the lung, the skeleton, the liver and the spleen in eight, five, one and ten patients, respectively.

The other 607 volumes were non-tumor tissues, including six composed NT+NT volumes. Most frequently, physiological skeletal uptake was segmented (n = 295), mainly located in spine and pelvis. Tracer uptake in the kidneys was also frequently segmented (n = 134), occasionally in multiple volumes per kidney due to uneven tracer distribution. 86 non-tumor volumes were located in the head-and-neck-region, mostly in the salivary glands, the Waldeyer´s ring or the vocal cord. Due to incorrect cropping, two urinary bladders were included in the further evaluation. An overview of all manually classified volumes is shown in the last line of Table 2.

Table 2

Confusion matrix of manual and automatic classification of the segmented volumes.

		Manual Classification
		T	T+NT	SK	HN	HT	RK	LK	LI	GI	G	BL	BF	NT+NT	sum
Automatic Classification	T	182	7	9	5	5	1	5	0	13	2	1	7	3	240
	T+NT	4	10	0	0	0	0	0	0	0	0	0	0	2	16
	SK	15	2	277	0	2	4	0	1	6	0	1	1	0	309
	HN	11	1	4	81	0	0	0	0	0	0	0	1	0	98
	HAT	1	0	0	0	24	0	0	0	0	0	0	0	0	25
	RK	2	0	1	0	0	61	0	0	0	0	0	0	0	64
	LK	6	1	1	0	0	0	60	0	0	0	0	1	0	69
	LI	0	0	0	0	0	0	0	0	0	0	0	0	0	0
	GI	3	0	3	0	0	1	2	0	17	1	0	0	1	28
	G	0	0	0	0	0	0	0	0	1	0	0	0	0	1
	BL	0	0	0	0	0	0	0	0	0	0	0	0	0	0
	BF	1	0	0	0	0	0	0	0	0	0	0	2	0	3
	NT+NT	0	0	0	0	0	0	0	0	0	0	0	0	0	0
	sum	225	21	295	86	31	67	67	1	37	3	2	12	6	853

Abbreviations: T—Tumor, T+NT—Composed volume of tumor and non-tumor tissue, SK—Skeleton, HN—Head-and-neck-region, HT—Heart, RT- Right kidney, LK—Left kidney, LI—Liver, GI—Gastrointestinal tract, G—Genital organs, BL—Urinary bladder, BF—Activated brown fat tissue, NT+NT—composed volumes of more than one non-tumor tissue type (NT+NT).

Automatic classification of tumor vs. non-tumor

Overall, 203/246 tumor lesions and 554/607 non-tumor volumes were classified correctly, corresponding to a sensitivity of 83%, a specificity of 91%, a PPV of 79%, a NPV of 93% and an f1-score of 81%.

43 tumor lesions were misclassified. These lesions were mistakenly considered to be non-tumor uptake, mainly in the skeleton (n = 17), the head-and-neck-region (n = 12) or the kidney (n = 9). 53 non-tumor volumes were misclassified as tumor lesions. Most of these misclassified lesions referred to non-malignant uptake in the gastrointestinal tract (n = 13), physiological uptake in the skeleton (n = 9) or uptake in activated brown fat tissue (n = 7) (Table 2).

Automatic classification of non-tumor tissues

The most common non-tumor tissues (skeleton, kidneys and soft tissue in the head-and-neck-region) were correctly classified in 94%, 90% and 94% of all cases. 77% of the heart volumes were correctly classified. The automatic classification of gastrointestinal uptake, activated brown fat tissue and composed NT+NT volumes was an issue. The rates for correct classification were 46% (17/37), 17% (2/12) and 0% (0/6), respectively.

Results on patient level

In 44/60 (73%) patients, all tumor lesions were correctly classified. In ten out of the 16 patients with misclassified tumor lesions, only one false-negative tumor lesion occurred. Misclassified tumor lesions were located in all body regions: in six patients in the neck, in four patients in the mediastinum and in six patients below the diaphragm. The algorithm missed splenic involvement in one patient and skeletal involvement in five patients.

Non-tumor uptake misclassified as tumor was seen in 31/60 (52%) patients, mainly below the diaphragm (n = 17). In 14/31 patients, only one non-tumor lesion was misclassified. In seven patients, physiological skeletal uptake was considered to be tumor.

In 13/37 patients with cardiac uptake, the heart was part of a composed volume. In 18 out of the other 24 patients (75%), the heart was correctly classified. In 3/6 patients with cardiac misclassification, the heart was untypically configured since only the valve plane showed increased tracer uptake Overall, in 22 patients all volumes were correctly classified. In nine patients both, misclassified tumor and non-tumor lesions were seen. An example for each type is shown in Figs 2 and 3.

Fig 2

Patient from the EuroNet-PHL-C1 study with an optimal automatic classification.

Fig 3

Patient from the EuroNet-PHL-C1 study with a suboptimal automatic classification and incorrect cropping at the upper border.

Discussion

Our automatic algorithm for segmentation and classification of lymphoma lesions in initial PET scans resulted in a sensitivity of 83% and a specificity of 91%. In 44 of our 60 patients, all tumor lesions were classified correctly. In ten patients only one tumor lesion was misclassified and in six patients more than one. Most frequently, tumor lesions were misclassified as non-malignant uptake in the skeleton or the head-and-neck-region.

Tumor lesions were discriminated from eleven different non-tumor uptake categories, including complex areas like salivary glands, skeleton and brown fat tissue. The classification of non-tumor uptake provided sensitivities ≥ 90% for the most common tissue types (skeleton, kidneys and soft tissue in the head-and-neck-region). These three tissues together amount to 85% of the number of all non-tumor volumes.

The automatic heart classification is challenging [23] due to the variable uptake behavior. In half of our patients with misclassified heart volumes an untypical cardiac uptake pattern was seen. However, 77% of all heart volumes were classified correctly.

The classification of non-malignant gastrointestinal uptake was difficult. Due to its focal appearance and variable localization, gastrointestinal uptake may appear like lymphoma lesions. In our study, more than half of all gastrointestinal volumes were misclassified, one third was mistakenly classified as tumor lesion.

Another critical category for automatic classification was activated brown fat tissue. Typically, it is characterized by multiple singular uptake areas mostly located in the neck and shoulder region. This area is also a preferred localization for lymphoma lesions. Thus, increased uptake in the brown fat tissue hampers the detection of lymphoma uptake, for physicians but especially for automatic algorithms.

Overall, in about half of our patients, at least one non-tumor volume was misclassified as tumor lesion.

For our algorithm the following conditions were chosen:

First, an SUV threshold of 2.5 for lesion segmentation was applied. This sensitive threshold [24, 25] was chosen to include all potential tumor lesions in our evaluation. Different segmentation thresholds were used by other authors: fixed thresholds of an SUV of 2.5 [12, 16], 3.0 [23] or 4.0 [26], or adaptive thresholds of 41%/42% of the SUVmax [11, 16, 26, 27] or background-related thresholds [12, 16, 21].

Second, a watershed algorithm was chosen for segmentation. Watershed segmentation is frequently used for tumor segmentation in PET imaging [20, 23].

Third, the z-axis of the PET scans was cropped for standardization of the scan length. Cropping was important in our multicenter approach since different PET acquisition protocols were used (e.g. skull base to mid-thigh, whole-body). As a result of automatic cropping, only very few lesions got lost.

Fourth, volumes below a size of 2.0 ml were not considered for classification. This approach was also applied by other authors and reduced the number of artifact-related lesions, prone to misclassification [12, 21]. Moreover, a lower threshold on volume size was also used for staging assessment in lymphoma trials [18, 28].

Fifth, a random forest was applied for classification of the segmented volumes. We used several hundred of trees and 31 features for the random forest. Feature standardization was achieved by transformation of the computed features to a standardized size and scanner resolution avoiding inaccuracies introduced by resampling.

Sixth, all tissue types with an increased uptake were included in our evaluation to address the situation in the daily routine. Overall, 14 different categories were used for classification. Most published studies only differentiated between tumor and non-tumor lesions [12, 27], while Hsu et al. [23] used a total of six classification categories.

Limitations

The main reason for incorrect automatic cropping was the absence of the brain on the PET scan. Our cropping algorithm should be improved regarding this special situation.

Composed volumes consisting of more than one tissue category are a major issue for automatic classification [23, 26]. In our study, 17/21 T+NT volumes and 5/6 NT+NT volumes were classified as tumor lesions. Thus, for our algorithm, composed volumes appeared to resemble tumor rather than non-tumor tissue, regardless of whether they contained tumor or not. Although T+NT volumes were classified as tumor these volumes did not exclusively represent tumor tissue. Thus, the assessment of tumor features could be skewed by the non-tumor parts. Automatic splitting of composed volumes into the individual tissues is not yet possible [26] but might be an option in the future. Manual splitting is a solution for individual cases but not for the assessment of large datasets. A potential approach for automatic splitting might be the organ segmentation in morphological imaging and the transfer of the segmented volume to the PET scan.

Comparison with other algorithms

A comparison of our results with published data is complicated by differences in methodology, tumor spread in included patients and inclusion of a varying number of non-tumor categories.

A similar methodology for lymphoma classification was published by Hsu et al. [23] using also PET data only, a watershed algorithm and a random forest. The authors identified tumor lesions with sensitivity of 84% and specificity of 92%. Compared to Hsu et al., we trained our algorithm on PET scans from multiple centers, used an automatic cropping algorithm for scan length normalization, included all tissue types with increased uptake in our classification and had more tumor lesions per scan (four vs. one). Interestingly, in our more complex setting we could confirm the very good performance of this methodology.

Several authors used convolutional neural networks (CNN) for automatic segmentation [11, 21] and/or classification [12, 27] of lymphoma lesions. Capobianco et al. [12] and Sibille et al. [27] presented a sensitivity and specificity of 80% and 88%, and of 75% and 96% for lymphoma classification. Both CNN studies used information from PET and corresponding morphological imaging. It should be noted that our algorithm achieved comparable results while only PET imaging and a significantly lower number of manually classified scans was used.

Conclusion

Our algorithm, trained on a small number of scans and on PET information only, showed a good performance and is suitable for automatic lymphoma classification. Compared to a CNN, a lower number of data is required for sufficient training of our algorithm.

The future method of choice for automatic lymphoma classification is still unclear. Different approaches should be developed and tested to find the optimal solution.

Automatic cropping algorithm.

The aim of heuristic cropping was to detect the brain and the urinary bladder as the cranial and caudal reference points. In most cases both show a distinct physiological uptake and therefore are a viable targets. In a first step the 3D SUV image (a) was projected onto the xz-plane (b), cropped to exclude extremities (c), and further projected onto the z-axis to create a function of SUV maxima from cranial to caudal (d). In a seconded step we smoothed the function and applied a peak detection to extract the first and last significant peak (e). Those correlated in most cases with brain and bladder and served as boundaries for cropping.

(TIF)

Click here for additional data file.

Explanation of the nested cross validation.

(DOCX)

Click here for additional data file.

8 Mar 2022

21 Mar 2022

Reviewer 1

Why did you only include patients with Hodgkin lymphoma? It seems interesting to add patients with other types of lymphoma mainly DLBCL.

You are right, the classification of patients with non-Hodgkin lymphoma would also be interesting and should be subject of a further study. It would be easily possible to use our method to train different classifiers in a group of DLBCL patients.

Our focus was to develop a classification algorithm for Hodgkin lymphoma patients. Our group is responsible for reference reading of PET scans in the EuroNet - pediatric Hodgkin lymphoma trials. Therefore, our aim is a further improvement of the prognostic information which can be derived from PET. Our next step will be the multifactorial search for prognostic factors for relapse in pretreatment PET of Hodgkin lymphoma patients. The presented automatic classification algorithm is a precondition for this study.

Sensitivity of the algorithm is too low (20% false negative). This is comparable to other algorithms, but hampers the ability to use the algorithm in a real life clinical setting. Please comment on this matter.

It is correct, a sensitivity of 83% probably hampers the use of our classifier in a real life setting, if the algorithm is used alone. However, our automatic algorithm could support the diagnostic work of physicians. Additional information from an automatic classifier will most likely improve the accuracy of the physicians report. Given the complex physiological FDG pattern and the variable occurrence of lymphoma involvement, our results seem to be quite solid, especially since only PET data was used in our study.

Most frequently, false negative lesions occurred in the skeleton and the head-and-neck area. The differentiation between malignant lymph nodes, inflammatory lymph nodes and accessory salivary glands in the neck can be sophisticated, even for experienced physicians. Likewise, the differentiation between physiological increased skeletal uptake and focal skeletal lesions may be also challenging for physicians. In an interreader study of five experts, discrepancies in reporting occurred predominantly in the same areas (Kluge et al. Inter-Reader Reliability of Early FDG-PET/CT Response Assessment Using the Deauville Scale after 2 Cycles of Intensive Chemotherapy (OEPA) in Hodgkin's Lymphoma. PLoS One. 2016;11:e0149072). An additional automatic analysis could help to harmonize the reporting.

Specifically, low specificity for gastrointestinal disease might be even more problematic with other types of aggressive lymphoma that tend to involve the GI tract more commonly than Hodgkin lymphoma (eg. DLBCL).

The correct classification of physiological gastrointestinal uptake was an issue in our study. However, none of our study patients had lymphoma involvement of the gastrointestinal tract. Only three patients had mesenterial lymph node lesions. Misclassification occurred between patchy intestinal uptake and abdominal lymph node involvement.

In lymphoma types with a tendency of GI involvement, GI lesions would occur frequently and often with large tumor manifestations in this region. An automatic classifier would probably learn how GI lesions look like and how to discriminate them better from non-malignant GI uptake. Thus, the algorithm performance for the classification of GI-lesions would most likely increase significantly.

Reviewer 2

The articles clearly states that different PET scanners of different centers were used to performe the study. In this setting, more information about the scanner used (PET or PET/CT, analogic or digital scanner, reconstruction algorithm..) are required, better if presented in a table.

We agree, it would be interesting to add an overview of the scanners used. Unfortunately, we are not able to present the data since patient data of all C1 patients were completely anonymized first, for data safety reasons. Thereafter, the 60 patients analyzed in our study were drawn randomly out of all C1 patients. Thus, it is not possible anymore to access and present the scanner data.

The "Automatic Classification" section even if really precise is hard to understand and to follow.

Thank you for this comment. We revised the “Automatic classification” section in the methods part. Now, a simplified explanation of the used method is given here. A detailed description can be found in the supplementary material.

Informations about tumor localization should be given: were all of them nodal localization or other organs were involved?

We added the following information in the “Manual classification” section in the result part:

“Tumor lesions occurred in lymph nodes above and below the diaphragm in 59 and 17 patients, respectively. Extra-nodal lesions occurred in the lung, the skeleton, the liver and the spleen in eight, five, one and ten patients, respectively.”

Submitted filename: Response to Reviewers_1.docx

Click here for additional data file.

6 Apr 2022

Automatic classification of lymphoma lesions in FDG-PET – Differentiation between tumor and non-tumor uptake

PONE-D-22-03090R1

Dear Dr. Thomas W Georgi ,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Domenico Albano

Academic Editor

PLOS ONE

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The manuscript is solid and intresting. The authors adressed my comments and explained why some of them are not in the scope of the cureent study.

Reviewer #2: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

8 Apr 2022

PONE-D-22-03090R1

Automatic classification of lymphoma lesions in FDG-PET – Differentiation between tumor and non-tumor uptake

Dear Dr. Georgi:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Domenico Albano

Academic Editor

PLOS ONE